Scientific Journal - Special Issue - November 2013

SCIENTIFIC JOURNAL OF THE TECHNICAL UNIVERSITY
OF CIVIL ENGINEERING
Mathematical Modelling
in Civil Engineering
BUCHAREST
Special Issue
November - 2013
Disclaimer
With respect to documents available from this journal neither U.T.C.B. nor any of its employees make any warranty,
express or implied, or assume any legal liability or responsibility for the accuracy, completeness, or usefulness of
any information, apparatus, product, or process disclosed.
Reference herein to any specific commercial products, process, or service by trade name, trademark, manufacturer,
or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring
by the U.T.C.B.
The views and opinions of authors expressed herein do not necessarily state or reflect those of U.T.C.B., and shall
not be used for advertising or product endorsement purposes
CONTENTS
PREFACE.................................................................................................................................................................... 5
CONSOLIDATION PROCEDURE FOR UNSATURATED SOILS USING A MODIFIED SETUP OF THE
TRIAXIAL COMPRESSION APPARATUS ........................................................................................................... 7
Adrian Liviu BUGEA
SLOPE STABILITY ANALYSIS OF UNSATURATED SOILS ......................................................................... 13
Andreea CARASTOIAN
EXPERIMENTAL AND NUMERICAL INVESTIGATION OF INDOOR COMFORT AND ENERGY
CONSUMPTION IN A TYPICAL ROMANIAN CLASSROOM FOR DIFFERENT GLAZING AREAS .... 23
Tiberiu CATALINA, Razvan POPESCU, Nicolae BAJENARU, Andrei ENE
INFLUENCE OF LONGITUDINAL VORTICES ON HEAT TRANSFER FOR AIRFLOW PASSING
THROUGH AN INNOVATIVE SOLAR FACADE .............................................................................................. 29
Cristiana Verona CROITORU, Florin BODE, Ilinca NASTASE
DESIGN PROCEDURE FOR SIDE WALLS OF SOCKET FOUNDATIONS .................................................. 38
Ionu DAMIAN
CONCEPTION OF AN ADVANCED THERMAL MANIKIN FOR THERMAL COMFORT ASSESSMENT
IN BUILDINGS AND VEHICLES.......................................................................................................................... 58
Angel DOGEANU, Ciprian CALIANU, George CHITARU, Matei GEORGESCU, Andrei TUDORACHE
ANALYSIS OF CREST CUTOFF WALL AT OSTROVUL MIC LEFT SIDE EMBANKMENT DAM ....... 69
Daniel Andrei GAFTOI, Ctlin POPESCU, Drago FRILESCU
ENERGY SAVING ANALYSIS INSIDE A DOUBLE SKIN FACADE ............................................................. 78
Sebastian HUDITEANU, Claudia-Florentina POENARI, Bogdan-Iulian BALINT, Monica CHERECHE,
Nelu-Cristian CHERECHE
OPEN SOURCE 3D MODELING FROM RASTER IMAGES ........................................................................... 84
Gabriel Adrian KEREKES
INFLUENCES OF OXIDATION STEP AND INITIAL METAL CONCENTRATIONS ON IRON AND
MANGANESE REMOVAL EFFICIENCY .......................................................................................................... 90
Alexandru JERCAN
SEISMIC RESPONSE OF TALL BUILDINGS WITH ROCKING WALLS SYSTEM ................................... 99
Lidia MARIN, Mircea VADUVA
NUMERICAL MODELLING OF RIGID VERTICAL INCLUSIONS AS REINFORCEMENTS FOR
COMPRESSIBLE SOILS..106
Iulia-Victoria TALPOS (NEAGOE)
MATHEMATICAL MODELLING OF SHOCK WAVES GENERATED BY BLAST EVENTS AND THEIR

EFFECT ON CONCRETE STRUCTURES ........................................................................................................ 113
George-Bogdan NICA
MATHEMATICAL MODELING FOR THE SEDIMENTATION PROCESS IN THE RESERVOIRS...... 122
Catalin POPESCU, Daniel Andrei GAFTOI, Drago FRILESCU
INTEGRATED SOLUTIONS FOR IMPROVING THE ENERGETIC PERFORMANCE OF BUILDINGS
.................................................................................................................................................................................. 131
Ana-Maria PASRE, Paul ANGHEL, Nelu-Cristian CHERECHE, Andrei BURLACU
R2D NUMERICAL MODELING OF FLOW THROUGH GEOMEMBRANE DEFECTS IN A LANDFILL
LINING SYSTEM .................................................................................................................................................. 137
Gheorghe PANTEL
NUMERICAL MODELLING OF PILED RAFT FOUNDATIONS ................................................................. 150
rpd SZERZ
INFLUENCE ASPECTS OF EXTERNAL PARAMETERS UPON EFFICIENCY OF A NATURAL SMOKE
EXHAUST SYSTEM.............................................................................................................................................. 159
Andrei-Mihai STOICA
DYNAMICS OF FREE SURFACE FLOW AROUND A CYLINDER VISUALIZATIONS AND PIV
MEASUREMENTS ................................................................................................................................................ 170
Nicoleta Octavia TNASE, Florin BODE
DYNAMIC BEHAVIOR OF BUCKLING RESTRAINED BRACES AND THE INFLUENCE OF
COMPRESSION STRENGTH ADJUSTMENT FACTOR .............................................................................. 176
Mircea VADUVA, Lidia MARIN
PREFACE
We are very glad to present this special volume of Mathematical Modelling in Civil Engineering
containing the best papers of the fourth edition of the Young Researchers Conference from the
Technical University of Civil Engineering of Bucharest. As in the case of past editions, this year
also, we received a considerable number of articles and we are proud to say that the scientific
quality of the papers is very high.
Keeping up the tradition of interesting Keynote Lectures, we are grateful to Professor Alex
Mahalov from the University of Arizona for the honour of accepting our invitation.
In the series of scientific workshops associated to our conference, we propose this year two
workshops: Exploratory research studies for Indoor Ambient Quality: methods, results and
perspectives and Environmental Quality in Hospitals: current problems, solutions and long term
strategies. Another scientific event that will take place during YRC 2013 edition will be an
information session regarding the Research and Innovation Programme Horizon 2020 organized
by the National Education Ministry and its national communication network.
This year the conference was held on the 21th and 22th of November 2013 at our University.
Financial support for organizing this event was provided by S.C. ERBAU EDIL CONSTRUCT
S.A.; Tensor SRL; SC METROPTICS S.R.L; CAMBI Research Center; SC ESRI Romania
SRL; SC CEMROM SA, which are gratefully acknowledged.
We want to recall our motivation in organizing this series of conferences dedicated to young
researchers. As we know, Romanian research has not always been privileged, and currently hard
times are announced for all those who contribute in this area and funds allocated to this area are
far from being increased. Perhaps the most difficult will be for the youngest of us who have not
won "independence" yet as confirmed researchers. These difficult times are accompanied in the
Romanian scientific community by the lack of communication and collaboration, although it is
known that modern research is not possible without collaboration.
The latter should take place both in the common areas and trans-disciplinary. Awareness of
problems and shortcomings of the scientific community should not scare us, rather we should be
encouraged to start making decision and changing things in order to get where we want. We, the
young researchers from UTCB, we believe this is in our power, to change things, and so we
decided to take the first step. The first step is to recognize that we are not satisfied with the
current situation and to propose a change.We want to be able to communicate and exchange
ideas at the level of other European research institutions. We also want to actively participate in
raising the visibility of our University and the community of young Romanian researchers. For
this reason, the first action that we want to take, in our desire to make our contribution to these
changes, was to organize the first Conference of the Young Researchers from the Technical
University of Civil Engineering of Bucharest. Or idea was to initiate a conference of young
researchers from our university open to other young engineering researchers. We believe that
finally came the time of positive discrimination organizing a youth conference for young people
to the standards we want. We also believe that communication is the key that will open new
opportunities for development based on cooperation and inter-disciplinary teams.
These series of events completes a lack in the scientific scenery of Romania. Our ambition was
to organize the first conference dedicated exclusively to young people (aged up to 35 years)
researchers in engineering and especially civil engineering. We wanted to give to our university
colleagues and to the colleagues from other universities a scientific framework to share their
concerns about their research subjects and their original results. We wanted them to realize that
Mathematical Modelling in Civil Engineering - Special Issue - 2013
we are a privileged community that works on passionate subjects since we all chosen to stay and
serve the Research, and that we can improve our condition by exchanging in a large scientific
network that will start to develop.
This special issue presents a selection of the best papers presented at the conference. All the
submitted papers in this special issue were fully peer reviewed by the reviewers drawn from the
scientific committee, external reviewers and editorial board depending on the subject matter of
the paper. Reviewing and initial selection were undertaken electronically. After the rigorous peer
review process, the submitted papers were selected on the basis of originality, significance, and
clarity for the purpose of the workshop. All the selected papers for the present special issue as
well as all the other contributions at our conference resulted in an exciting technical program for
which we want to warmly thank all the contributors.
The members of the scientific committee were:
Anton ANTON, UTCB, Bucharest, ROMANIA
Giorgio SERINO, Universita Degli Studi Di Napoli Federico II, Napoli, ITALY
Tiziana ROSSETTO, University College London, UK
Radu VACAREANU, UTCB, Bucharest, ROMANIA
Gabriel RACOVITEANU, UTCB, Bucharest, ROMANIA
Viorel BADESCU, UPB, Bucharest, ROMANIA
Loretta BATALI, UTCB, Bucharest, ROMANIA
Gabriel OPRISAN, UT Gh.Asachi, Jassy, ROMANIA
Alexandru ALDEA, UTCB, Bucharest, ROMANIA
Nelu - Cristian CHERECHES, UT Gh.Asachi, Jassy, ROMANIA
Ilinca NASTASE, UTCB, Bucharest, ROMANIA
Ionut - Ovidiu TOMA, UT Gh.Asachi, Jassy, ROMANIA
Alexandru DIMACHE, UTCB, Bucharest, ROMANIA
Vasilica CIOCAN, UT Gh.Asachi, Jassy, ROMANIA
Horatiu POPA, UTCB, Bucharest, ROMANIA
Victoria COTOROBAI , UT Gh. Asachi, Jassy, ROMANIA
Aurel SARACIN, UTCB, Bucharest, ROMANIA
Ionut RACANEL, UTCB, Bucharest, ROMANIA
Ion MIERLUS-MAZILU, UTCB, Bucharest, ROMANIA
Mihail Ioan SAVANIU, UTCB, Bucharest, ROMANIA
Tiberiu CATALINA, UTCB, Bucharest, ROMANIA
We would like to thank the scientific committee, the program chairs, the organizing committee
for their work.
We are grateful to all the participants and to all the persons that have contributed to a successful
first edition of our conference. We hope that our first goal which was making a first step in
developing a large scientific network of young engineering researchers was reached. We also
hope that all contributors and other interested parts beneficiated scientifically from this special
issue.
The young researchers from Technical University of Civil Engineering of Bucharest.
November 2013,
Bucharest, Romania.
Mathematical Modelling in Civil Engineering -Special Issue - 2013
CONSOLIDATION PROCEDURE FOR UNSATURATED SOILS USING A

MODIFIED SETUP OF THE TRIAXIAL COMPRESSION APPARATUS
Adrian Liviu BUGEA PhD Student, Technical University of Civil Engineering, Faculty of Hydrotechnical
Engineering, e-mail: livuadrian@yahoo.com
Abstract: This paper presents the proposed diffusion procedure for obtaining controlled suction
samples, in order to determine the soil water characteristic curve (SWCC). The method was
mentioned by D.G. Fredlund [1] and it was adapted to the triaxial compression equipment in the
geotechnical laboratory of the Technical University of Civil Engineering Bucharest. Based upon
the translation axis method, initially under a theoretical form, an optimal modified setup has been
designed. Moreover, a consolidation procedure permitting the determination of the volumetric
deformations is being analyzed, in order to obtain a volume variation law as a function of the
moisture content under an imposed value of suction.
Keywords: unsaturated soils, matric suction, volume variation law, SWCC
1. Introduction
Last decades trend in Soil Mechanics is represented by the understanding and modeling of the
unsaturated soil behavior, state in which over 80% of the cases of soil samples are found.
Moreover, this is of great interest as the direct foundation systems are founded on this type of
soil and certain particularities of their behavior are not covered by classical saturated soils
mechanics, among which we can specify collapsible or swelling-contractile properties of eolian
deposits or clayey layers.
Based on the works of researchers like D.G. Fredlund, H. Rahardjo [1], A. Gens or A. Lloret [2],
the knowledge accumulated in the field of agriculture has been transferred in the geotechnical
engineering, while from theoretical background new methods and apparatuses have been
developed ( [3], [4], [5], [6]). Initially using mercury or water burettes, in order to measure either
water pressures or pore air pressures, the nowadays technology implied in these tests evolved to
such a precision that almost infinitesimal fluid volumes may be measured or applied to the
samples.
This paper aims at presenting an evolved methodology to determine the constitutive surface of
soils, in terms of net stresses, void ratio and matric suction, using a classic triaxial compression
test apparatus, modified in order to be used in unsaturated soil tests. The sensors for monitoring
the sample during the test and the conditions applied to it in terms of cell pressure, back pressure
and pore air pressure are not new, but the entire setup and methodology have been designed in
such ways that it may offer a clear image on the behavior of the soil.
2. Variation of indices and their determination
Following the molding of the sample to a ratio between the diameter and the height of 1:2, for
which it is recommended to have a 50x100mm sample due to the stiffness reason, the sample is
measured and the initial mass and physical parameters are being identified. Based on the relation
of the general density with respect to the solid skeleton density, porosity and moisture content,
one obtains all the indices to define soils initial state.
During the actual consolidation test, following the setup of the sample in the modified triaxial
apparatus, a cell pressure, back pressure and air pore pressure combination is applied to the
sample until all the displacement and internal stresses are stabilized. At the end of the stage, the
void ratio, the matric suction and the net stresses that describe the new physic and mechanical
state of the sample are saved. This combination belongs to a surface that connects these three
components, describing the stresses that act on an unsaturated soil sample and its deformation
state. This surface is obtained using either odometer test monoaxial infinitely confined
compression test, or triaxial compression test. The Ct compression index is related to the at or av
compression index obtained using the oedometer test, determined using the saturated soil
mechanics relation:
(1)
where e is the variation of the void ratio and p the change in pressure. If the unloading case is
considered, the ats coefficient is obtained. The Cm index is determined considering volumetric
variation during a drying test.
Following this stage, new stress combinations may be applied, in order to obtain more points on
the net-e-(ua-uw) surface.
Fig. 1 - Constitutive surface of an unsaturated soil [1]
Although Fredlund and Rahardjo [1] implied that this surface may be determined by the aid of
odometer compression test and suction test using the translation of axes technique pressure
plate drying test, a new method that involves both wetting and drying paths can be developed. If
the air pore pressure is lowered, the newly obtained matric suction difference between air and
water pore pressures, is lowered, therefore, the soil sample will arrive to an equilibrium state by
adding water to its content the wetting path. If on the contrary, the air pressure is increased, the
water content in the sample will reach lower values.
3. Proposed setup of apparatus
Based on the aforementioned techniques, a modified triaxial test setup is being proposed, in
order to simplify the installation methodology of the sample and increase the accuracy of the
measured deformations.
Most of todays unsaturated triaxial compression apparatuses are using the double wall method,
in order to reduce the error regarding the deformations of the inner cell, affecting the volume
calculations of the fluid that creates the hydrostatical pressure on the sample, and, finally, the
calculations of the volume variations of the sample itself. This indirect method of determining
the volumetric variations of the soil is both subjected to errors and difficult to use due to its
complexity degree of setting up the sample inside the two celled apparatus.
Therefore, a more simple and direct setup has been developed, involving a dynamic triaxial
compression apparatus, designed initially for saturated soil mechanics. The main difference
between the saturated and unsaturated apparatuses are the presence of a high air entry value
8
ceramic stone, which can maintain a pressure difference between water and air and the
possibility to control another inflow fluid the air. Therefore, using the top cap drain, as shown
in Fig. 2, an air controller is linked to the upper part of the sample and through a porous stone the
air pore pressure is varied. In the pedestal, a ceramic stone is placed, in order to keep the
equilibrium between the pore air pressure (denoted with ua) and water back pressure (kept
constant and denoted with uw), without getting air into the water system. Moreover, the ceramic
disk is always held in a saturated state, due to its high capillarity and the water pressure applied
bottom-up.
Fig. 2 The proposed unsaturated single cell triaxial setup
In order to overcome the problem of varying cell volume, the direct method of using
displacement transducers to measure the radial and vertical deformations of the sample is being
proposed. As Fig. 2 shows, in the case of vertical displacement transducer, a LVDT (linear
variable differential transformer) is attached to the sample and its vertical deformation is
measured. Considering the case of isotropic consolidation, the loading ram is not touching the
top cap the undocked status, the vertical deformations cannot be determined but attaching the
local vertical displacement sensor. In the case of anisotropic K0 (coefficient of active earth
pressure at rest), consolidation step, although the distance between the porous stone outer
surfaces is known, as the initial value was declared to be equal to the initial height of the sample,
a more accurate displacement monitoring can be done using the attached sensor.
Although it is possible to back-calculate the radial deformation of the sample, knowing the cell
volume and the sample height variation, a more accurate method is to measure directly on the
sample the lateral surface variation - its radial deformation. In order to measure this type of
variation, another LVDT is placed on a ring attached to the sample (Fig. 2) and whose initial
position is set to a null value. All the measured linear deformations are treated as the perimeter
deformations of the cross section, taking into account that the expected displacement value is small.
As the tests are undergoing, diffused air may accumulate under the ceramic porous stone as
Fredlund [1] presented. Although several devices have been designed in order to determine the
volume of diffused air, such as the bubble pump [7], or even a simple system to flush the
diffused air beneath the ceramic disks [8], a newer technique developed by Lawrence [9] is to be
used. It consists of rapidly changing the pressure in the pore water line, in order to determine the
air volume using the ideal gas law. It still must be taken into consideration that these tests,
especially on clayey soils are running for a longer period several days to weeks, and the
diffused air volume may exceed the total volume of water of the tested sample. Therefore, a
flushing apparatus must be connected to the aforementioned setup, in order to relieve and keep at
low values the volume of the diffused air (Fig. 3).
Fig. 3 - The flushing device according to Fredlund [1]
The flushing of the diffused air is performed by opening the ball valve and moving to the tank a
sufficient amount of water which will be carrying along the air bubbles. The water level is
brought back to the initial level by the pressure and volume water controller. The difference
between the initial position and the new one offers the volume of diffused air in terms of
10
pressure for the differential pressure transducer, which measures the height of the water column
in the flushing device.
Moreover, several authors ( [1], [10], [11]) also emphasized the possibility of wetting by
accident the sample if the system is using water as cell pressure fluid. Therefore, a double
membrane system with a film of oil between and a layer of silicone grease on the top cap and the
pedestal are recommended to be used, adding to the mechanical waterproofing acquired by
utilizing the O-rings.
4. Methodology of the consolidation test
In order to determine the constitutive surface of an unsaturated soil, as presented in Fig. 1, a
series of multiple stages are to be taken into consideration. One must bear in mind that both of
the wetting and drying path may be used, so a planning of the test is needed, as the well-known
hysteretic behavior of the soil with respect to wetting and drying modifies its properties (Fig. 4).
Fig. 4 - Suction curve and hysteretic behaviour of the drying and wetting paths
If applying a drying path is considered, during the consolidation stage, a higher air pore pressure
is used. At the equilibrium between the matric suction and the moisture content of the sample, all
the deformations will reach a constant value, obtaining one of the points of the constitutive surface.
In the case of considering a wetting path, a higher pressure is applied to the pore water line, in order
to impose a gradient that will reduce the suction of the sample. In both cases, it is recommended to
know the soil water characteristic curve (SWCC), in order to determine the value of the moisture
content at the final moment, according to the detected matric suction (ua-uw).
The following steps and getting more points respectively on the surface are obtained by
modifying the cell pressure, increasing its value. Each step may contain various suction values,
in order to determine the form of the surface in all three separate coordinate systems: net stress
(-ua) matric suction (ua-uw), void ration (e) matric suction (ua-uw) and net stress (-ua)
void ratio (e). Composing these three curves into one three-dimensional surface, offers the
constitutive relations used in modeling the behavior of the unsaturated soils, using mathematical
models such as Barcelona Basic Model [12].
11
Adding to these results of the consolidation tests the radial and vertical displacements also may
be plotted in different systems, with respect to the matric suction, applied stresses and moisture
content or void ratio, in the end, offering a more complex view on the behavior of the
unsaturated soils, especially used for the case of predicting the swelling and contraction
properties of soils ([13], [14]).
If the case of determining the shear resistance and the dynamic parameters with respect to the
soil matric suction, the considered setup may be used, but the results and the methodology is still
submitted to discussions.
5. Conclusions
This paper presented a setup to modify a classic dynamic triaxial compression apparatus in order
to determine unsaturated soil consolidation surface, in terms of applied net stresses, matric
suction and void ratio.
The aim of this new setup is to overcome some of the difficulties presented by other authors and
obtaining optimum installation, testing procedures and conditions in terms of device complexity
and sample monitoring, as well as accuracy of results.
Although at first it has been designed bearing in mind only the monitoring of the swellingcontractive behavior of clayey soils under constant stress state (consolidation step), further
research in terms of determining the shear resistance and even dynamic parameters of
unsaturated soils is developed, aiming at offering a clearer and more precise view on the entire
response of the soil material under different types of stresses that may lead to failure.
References
[1] Fredlund D., Rahardjo H., 1993, Soil mechanics for unsaturated soils, New York: John Wiley & Sons Ltd..
[2] Romero E., Gens A., Lloret A., 2001, Temperature effects on the hydraulic behaviour of an unsaturated clay,
Geotechnical and Geological Engineering, nr. 19, pp. 311-322.
[3] Padilla J., Houston W., Lawrence C., Fredlund D., Houston S., Perez N., 2006, An automated triaxial testing
device for unsaturated for unsaturated soils, 4th International Conference on Unsaturated Soils, Arizona,.
[4] Nishimura T., Fredlund D., 2003 A new triaxial apparatus for high total suction using relative humidity, 12th
Asian Regional Conference on Soil Mechanics and Geotechnical Engineering, Singapore.
[5] Lauer C., Engel J., A triaxial device for unsaturated sand - New developments, Dresden.
[6] Ho D. , Fredlund D., March-June 1982, A multistage triaxial test for unsaturated soils, Geotechnical Testing
Journal , vol. 5, pp. 18-25,.
[7] Bishop A., Donald I.,1961, The experimental study of partially saturated soils in the triaxial apparatus,
Proceedings of the 5th International Conference on Soil Mechanics and Foundation Engineering, Paris,.
[8] Padilla J., Perera Y., Houston W., Fredlund D., Perez P., 2006 Quantification of air diffusion through high-air
entry value ceramic disks, Proceedings of the 4th International Conference on Unsaturated Soils, Carefree,.
[9] C. Lawrence, 2005, Pressure pulse technique for measuring diffused air volume, Proceedings of the
International Symposium on Advanced Experimental Unsaturated Soil Mechanics, Trento,.
[10] Rahardjo, H., Fredlund D., December 1996 Consolidation apparatus for testing unsaturated soils,
Geotechnical Testing Journal, vol. 19, nr. 4, pp. 341-353,.
[11] Wulfsohn, D., 1994, Triaxial testing of unsaturated agricultural soils, n International Summer Meeting,
Kansas City,.
[12] Alonso E., Gens A., Josa, A., 1990 A constitutive model for partially saturated soils, Geotechnique, vol. 40,
nr. 3, pp. 405-430,.
[13] Hung Q., Fredlund D., 2004, The prediction of one-, two- and three dimensional heave in expansive soils,
Canadian Geotechnical Journal, vol. 41, nr. 4, pp. 713-737,.
[14] Fredlund D., 1975 Prediction of heave in unsaturated soils, n 5th Regional Conference, Banglore,.
12
SLOPE STABILITY ANALYSIS OF UNSATURATED SOILS

Andreea CARASTOIAN PhD student Technical University of Civil Engineering, Bucharest,
e-mail: carastoian.andreea@yahoo.com ;
Abstract: The paper presents the slope stability analysis of unsaturated soils considering
geotechnical parameters and groundwater level variations due to saturation degree. Infiltrations are
one of the main factors causing slope failures. The main parameters associated with slope stability
analysis are the characteristics of water flow, change of pore-water pressure and shear strength of
soils. The saturation degree of soils highly influences the geotechnical parameters. The finite
element method was used to evaluate the stability of the slope.
Keywords: slope analysis, unsaturated soil, groundwater, finite element method, geotechnical
parameters
1. Introduction
The stability factor of a slope can be computed using the finite element method by reducing the
soil strength until the slope fails. The resulting stability factor is the ratio of the actual shear
strength of the to the reduced shear strength at failure. In total stress analysis of soil slopes, total
stress shear strength parameters ( and ) are often used. Pore pressures are not considered.
These total stress analyses are appropriate in the short term only and not in the long term where
slope stability is a minimum (Simons et al., 2001). [1]
It is important to formulate a slope stability analysis method, which can track the failure process
from the initial deformation to the ultimate failure
2. Slope Stability Analysis
The slope stability analysis is an analytical tool for assessing the stability of a slope by using a
simple failure model in the analysis.
The required safety factor depends on the consequences of losses in terms of property, lives and
cost of repair in the event of slope failure. The safety factor is also dependent on the reliability of
design parameters.
A factor of safety is placed on shear strength parameters;
-
The strength parameters are independent of stress-strain behavior;
Some or all of the equations of equilibrium are used to the determine the safety factor;
Forces involved in the equilibrium methods are statically indeterminate.
2.1. Stress Analysis Method (SAM)

The failure of soil slopes came in a wide variety of conceivable manners. The qualitative
definition is given in a book by Terzaghi, Peck and Mesri (1996): The failure of a mass of soil
located beneath a slope is called a slide. It involves a downward and outward movement of the
entire mass of soil participating in the failure. [2]
Slope stability analysis is globally performed by means of two methods: the simplified method
and the numerical analysis method. The limit equilibrium which belongs to the simplified
methods is one way to approach slope stability problems and is most widely used in practice.
However, this method cannot consider the stress hysteresis effects during the formation of slope
and the variation of stress for foundations due to the groundwater. Although the finite element
techniques considers the formation process of slope and the property of foundation, the method
13
requires high cost and long analysis time, and the slope stability data for evaluation purposes is
relatively insufficient.
In this method, stress analysis is first performed on the slope using the finite element method.
Based on the stress analysis results, the factor of safety for potential sliding surface is calculated,
and the critical section is determined using the minimum safety factor. [3]
The safety factor in the finite element method is generally defined as:
(1)
Where,
is the shear stress, and
is the shear strength according to Mohr-Coulomb failure
criteria.
Fig. 1 - Stress components of sliding surface [4]
The initial failures for most of the unsaturated soil slopes have small depth-to-length ratios and
form the failure planes parallel to the slope surface; hence, the use of infinite slope analysis for
stability evaluation is thus justified (Collins and Znidarcic, 2004). The factor of safety of the
slope is calculated by using a modified Mohr-Coulomb failure criterion (Fredlund et al., 1978;
Fredlund and Rahardjo, 1993):
FOS
tan
tan
(2)
(3)
Where, c is the effective cohesion,

is the effective frictional angle,
normal stress,
is the pore-air pressure,
is the pore-water pressure,
is the internal friction angle due to matric suction.
suction,
is the net
is the matric
The unsaturated friction angle ( ) depicts the increment rate of shear strength due to an increase
in suction and it can be obtained by performing a series of triaxial compression tests under
various matric suction conditions. In these tests the pore air pressure ( ) control and transducer
are installed to measure the matric suction
. For finite slope analysis, the factor of
safety (FOS) of an unsaturated slope is expressed as:
where W is the weight of slice, which is the product of
the assumed slip surface) and is the slope angle.
(the total unit weight) and
(vertical depth of
The limit equilibrium methods calculate a factor of safety which, by definition, is assumed to be
the same at all points along the potential slip surface. This is reasonable only at failure, when all
the slices are on verge of failure; that is, when the factor of safety equals unity for each slice. In
reality, the local factors of safety will vary somewhat along the slip surface, for some slices it
might be one and higher value for others. In the case of brittle materials, even small hydrostatic
loads can reduce the local factors of safety to less than unity and the progressive failure
mechanism may be triggered. This happens, for instance, in over-consolidated clay that can
exhibit residual shear strength under drained loading and in loose, saturated sands under
14
undrained loading. Therefore, the methods cannot explicitly model the mechanism of progressive
failure (Pyke, 1991).[4]
Modeling of slope failure by the finite element method must address the following issues:
-
Occurrence of large deformations;
The effect of three dimensional conditions;
Accurate following of the equilibrium path under hardening, softening and snap-back
behavior;
Occurrence of narrow, continuous localized zone on which the slope slides.
In total stress analysis, the pore water pressure within the soils is ignored, only the free water
body is considered. Free water can be modeled as a material with unit weight,
but has no
strength (Figure 2a) or can also be modeled as an equivalent loading on the slope (Figure 2b).
Since the free water body acts as counter weight, a submerged slope is always more stable than a
dry one in total stress analysis.
Fig. 2 - Total stress analysis of the slope [4]
2.2. Strength Reduction Method (SRM)

The finite element method is a precise numerical analysis method which satisfies the force
equilibrium, compatibility condition, constitutive equation and boundary condition at each point
of a slope. It simulates the actual slope failure mechanism and determines both the minimum
factor of safety and the failure behavior. It can also reflect real in-situ conditions better than most
methods. Moreover, it can determine the failure process without assuming any failure planes in
advance. (Griffith 1999; Matsui, 1990).[5]
There are two types of methods used in the finite element method to analyze slope stability
Strength Reduction method and Indirect method. The Strength Reduction method is a direct
of sloped ground materials until failure
method, which gradually reduces the shear strength
takes place. Failure is assumed to occur when the analysis does not converge. Under this
condition, the maximum strength reduction factor where the analysis fails to converge becomes
the factor of safety. The Indirect method determines the factor of safety by applying the
calculated stresses combined with the conventional Limit theory method. The Strength
Reduction method uses a finite element technique first proposed by Zienkiewicz (1975). We now
focus on a Gauss point, A, of an element in a sloped ground structure to calculate the factor of
safety of a slope as shown in Figure 3. The stress state at this point is represented in a Mohr
circle. In order to represent the sliding surface, the shear stress at the point is divided by a factor
of safety, F, so that the Mohr circle for the stress state of the fictitious sliding surface becomes
tangent to the failure criterion.
15
Fig. 3 - Strength Reduction Method [6]
That is, the stress state of the point is corrected to the failure state. An increase in the number of
points results in a global slope failure. As soon as a finite element solution diverges, the analysis
stops and the limit value, F, becomes the minimum factor of safety for the slope. This method
requires stability in the numerical analysis, but returns consistent results and evaluates the actual
failure behavior. [6]
To determine the minimum stability factor of the slopes, the elasticity modulus
and Poissons
ratio
are assumed to be constant. The cohesion
and friction angle
are simultaneously
reduced, and the factor of safety, is determined at the diverging point. The factor of safety for
slope failure is determined on the basis of shear failure as follows:
(4)
Where,
is the shear strength of the sloped material.
This is computed according to the Mohr-Coulomb criterion.

(5)
Where,
- Coefficient of shear strength;
- Coefficient of shear strength;
SRF Strength reduction failure.
causing the
In order to determine the SRF accurately, it is necessary to trace the resulting values of
slope to fail. The incremental parameter is increased in very small steps even though it may extend the
analysis time duration. Otherwise, calculating the minimum factor of safety may face difficulties.
3. Example studies
3.1. Studied cases
The finite element stability analyses were performed on homogeneous slopes. Two sets of runs
were performed on each slope. The first case was a homogeneous slope consisting of the two soil
types considering the influence of a horizontal water table at various depths, such as 3m, 5m, 7m
and 10m. In the second set, on the same geometry, I analyzed the influence of the variation of the
shear strength parameters. For both sets I performed the slope stability analysis using stress
analysis method (SAM) in 2D and shear reduction method in 3D (SRM). The analyses were
performed using the finite element software Midas GTS.
Table 1 indicates the parameters for the finite element slope stability analysis for the first case,
while Table 2 shows the corresponding parameters for the second case.
16
Table 1.
Table 2.
Soil parameters for case 1
Soil parameters for case 2
Soil type
Case 1 - Soil 1
30
20
Case 1 Soil 2
10
20
20
30
25
10
20
15
15
20
The other parameters such as Youngs modulus and Poissons ratio are given the nominal values
and
, in both cases. The slope geometry and its finite element mesh are
of
shown figure 4.
Fig. 4 - Geometry and finite element mesh
3.2. Numerical Results for First Case

The numerical results of the safety factor are compared to each method, SAM and SRM and are shown in
Figures 5-14.
2D
results
using
SAM
method
without
water
table.
FOS
1,706.
Fig. 5 - a) Displacements contour of the slope; b) Shear strength of the slope.
2D results using SRM method without water table. FOS = 1,30.

17
2D results using SAM method with water table at 3 m depth. FOS = 1,62.
2D results using SRM method with water table at 3 m depth. FOS = 0,975.

18
We can observe the difference of the value of the safety factor using those finite element methods. Using
the stress analysis method, the value of the safety factor is bigger than 1.2, which means that it is stable.
In the other method, by reducing the shear strength parameters, the program finds the most unfavorable
situation, reducing the value of the safety factor to slope instability.
The next figures show the comparison of FOS with the shear strength parameters, influenced by the
fluctuation of water table.
Fig. 15 16 - FOS values versus shear strength parameters, influenced by fluctuation of water table.
19
Numerical result for slope analysis soil 2

Following the same steps the results for soil 2 are presented in the figure 16.
We observe the influence of the shear strength parameter modification. The initial value of safety
factor (FOS = 1.706) is lesser than the corresponding value (FOS = 1.38) for soil 1. Applying the
strength reduction method, the safety factor remains in the unstable zone.
3.3. Numerical Results for the Second Case
In the second study I present a comparison between 2D and 3D analyses. The hypothesis is to
observe the evolution of the safety factor by modifying the shear strength parameters
consecutively, like in table 2.
-
2D and 3D results using SAM and SRM method. FOS (SAM) = 1,706. FOS (SRM) = 1.2.
and
Fig. 17 - a) 2D - Displacements contour of the slope; b) 2D - Shear strength of the slope - SAM
Fig. 18 - a) 3D - Displacements contour of the slope; b) 3D - Shear strength of the slope SRM
2D and 3D results using SAM and SRM method. FOS (SAM) = 1,64. FOS (SRM) = 1.2.
[1].
and
Fig. 19 - a) 2D - Displacements contour of the slope; b) 2D - Shear strength of the slope SAM
20
2D and 3D results using SAM and SRM method. FOS (SAM) = 1,59. FOS (SRM) =
1.05.
and
2D and 3D results using SAM and SRM method. FOS (SAM) = 1,53. FOS (SRM) =
0.95.
and
The results are shown in the next figure, comparing the safety factor with the shear strength
parameters.
21
Fig. 25 - Comparison of FOS on SAM and SRM methods
In this case, we can observe the difference between those two types of analyses. In 2D, applying the SAM
method we can observe that the value of the stability factor is in the stable zone, modifying in each steps
the shear strength parameters. With the same condition, applying the strength reduction method, in 3D,
the value of stability factor is calculated using the reduced strength, being in the unstable zone.
4. Conclusions
The paper presents a comparison of two methods the stress analysis method and the strength reduction
method using the finite element method for analyzing the slope stability for an unsaturated case.
The slope stability analysis of unsaturated ground requires to simultaneously computing deformation and
groundwater flow with time dependent boundary conditions.
According to the results we can observe the advantages of the second method. SRM slope analysis can
produce insights into the failure mechanisms, and their formation, in ways that may not be as evident in
limit-equilibrium analysis.
The factor of safety is calculated using reduced strength, and the critical cross section is the area where
the maximum shear strain occurs.
Slope stability analysis of unsaturated ground requires to simultaneously computing deformation and
groundwater flow with time dependent boundary conditions.
References
[1] Simons, N.E, Menzies, B.K. and Matthews, M.C., 2001 - A Short Course in Soil and Rock Slope Engineering,
Thomas Telford, London.
[2] Terzaghi, K., Peck, R., B., and Mesri, G., 1996 - Soil mechanics in engineering practice, third ed. John Wiley &
Sons, Inc.
[3] Charles, W., W., Ng., Bruce, M., 2007- Advanced Unsaturated Soil Mechanics and Engineering, Taylor &
Francis Group, New York;
[4] Pyke, R., M., 1991 - TSLOPE Users Guide, TAGA Engineering Systems & Software Lafayette, California, ()
[5] Griffiths, D.V., and P.A. Lane, 1999- Slope Stability Analysis by Finite Elements, Geotechnique, vol. 49, no.
3, pp. 387-403.
[6] Shukra, R. and Baker, R., 2003 - Mesh Geometry Effects on Slope Stability Calculation by FLAC Strength
Reduction Method Linear and Non-Linear Failure Criteria. In Proceedings of the 3rd International FLAC
Symposium, Sudbury, Ontario, Canada, eds. R. Brummer et al, pp. 109-116.
[7] *** MIDAS GTS SOFTWARE, 2012, User Guide.
22
EXPERIMENTAL AND NUMERICAL INVESTIGATION OF INDOOR

COMFORT AND ENERGY CONSUMPTION IN A TYPICAL ROMANIAN
CLASSROOM FOR DIFFERENT GLAZING AREAS
Tiberiu CATALINA Lecturer, PhD, Technical University of Civil Engineering - Romania, Faculty of Building
Services, e-mail: tiberiu.catalina@gmail.com
Razvan POPESCU - Lecturer, PhD, Technical University of Civil Engineering - Romania, Faculty of Building
Services, e-mail: razvan22@yahoo.com
Nicolae BAJENARU Eng., Technical University of Civil Engineering - Romania, Faculty of Building Services,
e-mail: bajenaru.nicolae@gmail.com
Andrei ENE - Student, Technical University of Civil Engineering - Romania, Faculty of Building Services,
e-mail: eneandrei13@gmail.com
Abstract: This article is divided in two parts: in the first part the indoor conditions in a typical
classroom using experimental measurements are illustrated and in the second part, using numerical
simulations, the impact of glazing area on the energy consumption for heating/lighting and indoor
comfort is analyzed. It was found that in this typical classroom the indoor levels of CO2 do not
fulfill the requirements for a good air quality. However, thermal comfort is achieved as the air
temperature is around 20oC. Moreover, the illuminance level is also achieved as the window area is
around 25% from the floor area. During the numerical simulations we have studied a similar
classroom with the one from the experimental campaign. Four windows-to-floor-area-ratios have
been studied in terms of energy consumption, thermal and visual comfort. It has been found that a
larger area of window affects greatly the cooling demand, but has a slight impact on the operative
temperature. Moreover, the energy needed by the artificial lighting is reduced by 51% if we
compare a window-to-floor-ratio (WFR) of 15% to WFR 30%. It is recommended not to overpass a
WFR of 25% as this will affect too much the energy consumption.
Keywords: experimental measurements, indoor environmental quality, numerical simulations,
energy consumption, design optimization
1. Introduction
The reduction of energy consumption is an important point on the agenda of objectives to be
attained by EU countries by year 2020. At the same time, we cannot neglect the true purpose of
a building: to provide the occupants a comfortable and healthy indoor environment. Indoor
environmental quality (IEQ) is a concept that deals not only with thermal conditions but it also
goes much further, because it involves air quality, lighting and acoustics [1]. Given the many
interactions between building energy performance and IEQ, these two issues must be
addressed and researched in a connected manner. IEQ and energy are closely linked and only
an integrated research project can ensure that improvements in energy efficiency do not reduce
IEQ, and that improvements in IEQ do not decrease energy efficiency, either. This issue is
clearly expressed even by the European Energy Performance of Buildings Directive
2002/92/EC which underlines that the expression of a judgment about the energy consumption
of a building should be always combined with an analysis of the IEQ. The problem is even
more complicated as these are conflicting criteria [2] and finding an optimal solution can be
quite difficult for the current knowledge frontier.
Considering this panorama, it is highly important to search for solutions of building design that
provide the highest benefit for both energy saving and IEQ at the lowest cost. Educational
facilities are among the most important fixtures in our community [3], where children spend
around 25% of their time inside school classrooms, this area being like their second home [4].
As schools present a much higher occupancy than any other building, it is vital to have an
indoor climate that will not affect the comfort, health or intellectual performance of the
students [5]. In all the existing educational facilities in Romania many complaints were made
about discomfort, poor air quality (health problems) and expensive bills for facility operation.
This research project proposal emphasizes research on schools because these types of buildings
23
have been underrepresented in prior research. Moreover, this category was chosen to be
studied because children have a greater susceptibility to some environmental parameters than
adults [6] and there are limited or incomplete research studies in this area. Reported findings
on the indoor air quality (IAQ) in schools [7-8], correlation between IAQ and energy [9-10] or
the use of alternative ventilation systems in schools [11] have been made during the last
decade. Provision of useful information on the frequency of exposure to indoor factors in the
educational environment with precise survey data and experimental measurements is more than
needed. In this article we tackle this important aspect and also focus on the impact of glazing
area or lighting control on the energy demand and visual comfort. Finally, the study is
completed with a full simulation of thermal comfort along with energy consumption for
heating.
2. Experimental campaign
The experimental measurements took place in a classroom from a school situated in Rmnicu
Vlcea region, more exactly in Mateeti which is a rural area. The investigation of indoor
comfort was done on 1st October 2013 for the entire occupational period of that day, more
specific from 08:00 am to 12:00 am. The equipment that was used is TESTO 480 that is a multifunction tool that can record the following parameters: air temperature, globe temperature,
illuminance, air velocity, relative humidity and CO2 level. Moreover we have used the
equipment Dylos 1100 Pro Air Quality for dust particles counting for 0.5 and 2.5 microns. For
the sound measurements we used a sound meter class 2 precision. We have also recorded the
outside CO2 level, the relative humidity and air temperature for the same period. The equipment
used was a data logger type CO2 Meter mounted outside the school.
Fig.1 - Photos of the TESTO 480 installed in the classroom
The measurements were done with a time step of 1 min for the entire period. The room size is
8.7 m x 5.7 m x 3.48 m (L x l x h). During the measurement the room was occupied by 15 pupils
and 1 professor. Three windows with a surface of 3.1 m2 each allow the daylight to enter into the
occupied space. The windows are double pane glazing with a PVC frame and are considered a
typical solution to replace the older windows of the school. The floor is covered with linoleum,
while the rest of the walls and the ceiling are made of concrete painted in white. Three radiators
with of length of 1.8 m and 0.6 m the height are installed in the room. These radiators are sized
in order to ensure a good thermal comfort inside the room.
24
CO2 level [ppm]
Illuminance level [lux]
Temp level [oC]
RH level [%]
3500
90,0
26,0
3000
85,0
24,0
1050
950
Indoor
air temperature
850
80,0
2500
22,0
Outdoor
humidity
750
75,0
20,0
70,0
18,0
650
2000
1500
1000
Indoor
concentration
65,0
Outdoor
concentration
60,0
14,0
55,0
12,0
50,0
7:26
10,0
7:26
500
0
7:26
8:45
10:04
11:24
Time (hh:mm:ss)
12:43
8:45
10:04
11:24
Time (hh:mm:ss)
Heating ON
550
450
16,0
Indoor
humidity
12:43
Indoor
illuminance
350
Outdoor
air temperature
250
150
8:45
10:04
11:24
Time (hh:mm:ss)
12:43
50
7:26
8:45
10:04
11:24
Time (hh:mm:ss)
12:43
Fig.2 - Indoor comfort and air quality variation inside the classroom
From Figure 2 it can be noticed that the CO2 level before the pupils arrive to school was situated
at 600 ppm and then increased up to 2800 ppm. The indoor air quality norms are not fulfilled as
the maximum allowed value should be around 800 ppm, which is a translation of a poor air
quality. The explanation for this high values is that the classroom does not have a ventilation
system to supply the fresh air and that the windows have a good air tightness. This lack of fresh
air is translated in a low energy consumption zone, but at the same time in a low air quality. Due
to heavy rains the day before the measurement we have recorded a high value for the relative
humidity, while the indoor levels are in the range of 55% to 60%, more than acceptable. The
outside temperature during that period was low with values around 11oC which triggered the
heating system to start but only for 2 hours.
Due to building thermal inertia and heat gains from the pupils the indoor air temperature was
kept in a comfortable range of 20oC to 22oC. The indoor illuminance was also measured and the
mean value for the entire period overpasses the visual comfort standards [12]. It must be
mentioned that during the measurements the sky was covered by thick clouds and the artificial
lighting was turned on for the school hours. It can be concluded from this experimental
investigation that the building performs well in terms of thermal and visual comfort but has a
serious problem with the indoor air quality. Moreover, it was observed that the artificial lighting
control is not optimized and there is a waste of electric energy.
As noise problems are concerned, it was found that the global sound pressure level was higher
(45 dB(A)) than 35 dB(A) which is the limit value for a comfortable environment. This is due to
the sound that is propagated from the outdoor to the indoor space mainly through the windows.
The exterior environment was found to be noisy as the traffic is intense and multiple trucks are
passing by the school faade.
3. Numerical simulations
After the experimental analysis we have done a theoretical study of a classroom that has the size
of 8m x 6m x 3.2 m (L x l x h) using numerical simulations. We found these dimensions relevant
as most of our studied classrooms had these sizes. One of the most trustworthy solutions to
evaluate the indoor conditions and energy consumption of a building are the dynamic
simulations. It is clear that during the design stage of a new building it is important to verify the
impact of a certain parameter on the building performance. As the building faade is a key
element for the indoor conditions we wanted to study the impact of the window area on the
thermal comfort and energy consumption. Like for the experimental studied classroom this
simulated virtual classroom has a good air tightness of 0.2 ach/h and the installed windows are
double pane glazing windows. The walls are made of brick and are insulated with 7.5 cm of
25
polystyrene. The structure of the walls, the floor and the ceiling corresponds to real schools. The
simulated classroom is oriented WEST and is situated at the last floor of a building in order to
have the most disadvantageous case from the thermal point of view. The ceiling is adjacent with
an attic insulated with 10 cm of mineral wool.
3500
Heating demand
Energy [kWh/an]
3000
Cooling demand
Artificial lighting
2500
2000
1500
1000
500
0
WFR 15%
WFR 20%
WFR 25%
WFR 30%
Fig.3 - Annual heating, cooling and electric energy consumption
The simulations were done for the entire school year (15 September to 15 June) and we have
taken into account the occupation periods with the holidays/weekends/days hours, the heat gain
from artificial lighting and from pupils. The time step of the simulations was set to 1 hour and
they were realized using the TRNSYS 16.0 software.
The same scenarios of window-to-floor-ratios were used to investigate the impact on the heating
and cooling demand during the coldest and warmest day of the school year (see Figure 4). If for
the heating we do not observe large variations for the cooling demand we have a much larger
difference of up to 1500 W between WFR 15% case and WFR 30% scenario.
Cooling demand[W]
Heating demand[W]
Like it was mentioned previously the purpose of the numerical study was to check the building
performance for different glazing areas. We have considered four cases of window-to-floor-ratio
(WFR): 15%, 20%, 25% and 30%. From Figure 3 it can be noticed that there is an increase of 24
% for the heating demand between the WFR 15% case and WFR 30%. At the same time there is
an increase of 51.4% of the cooling demand for the same comparison. On the other hand we
observed a reduction by 43% of the electric energy if we increase the WFR from 15% to 30%.
Heating demand in the coldest day of the school year
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
WFR15%
Schoolhours
WFR20%
WFR25%
WFR30%
10
10
12
14
16
18
20
22
16
18
20
22
Cooling demand for the hottest day of the school year
5000
4000
3000
2000
1000
Occupation period
12
Hour[h]
14
Fig.4 - Heating and cooling demand for the coldest/warmest day of the school year
Thermal comfort is defined by ASHRAE [13] as that condition of mind which expresses
satisfaction with the thermal environment and is assessed by subjective evaluation. In our case we
express the thermal comfort using the operative temperature which translates the influence of both
air temperature and radiant surrounding temperature. The set point temperature for the air was
considered of 20oC and it can be observed that during the coldest day the operative temperature is
lower than 20oC, but also equal for the WFR scenarios. For the summer period the differences
between the four scenarios of WFR are higher but are still low with values around 0.5oC.
26
Operative temperature [oC] Coldest day
30,0
19
18
17
Occupation
period
16
WFR15%
WFR20%
WFR25%
WFR30%
29,5
WFR15%
WFR20%
WFR25%
WFR30%
Temperature [C]
Temperature [C]
21
20
Operative temperature [oC] Warmest day
29,0
28,5
28,0
27,5
27,0
26,5
26,0
Occupation
period
25,5
15
0
25,0
8 10 12 14 16 18 20 22
Hour[h]
8 10 12 14 16 18 20 22
Hour[h]
Fig. 4 - Heating and cooling demand for the coldest/warmest day of the school year
The illuminance level was obtained by using DIALUX software and several simulations were
necessary to have the data for different outdoor illuminance. Once the correlation between indoor
levels and outdoor illuminance was obtained, then by using a regression equation we were able to
calculate for each hour of the school year what the indoor illuminance level was. If the 300 lux level
was not attained then the artificial lighting was turned on and the energy consumption would be of
872 W. The simulations were conducted for an overcast sky, as mentioned by Comission
Internationale de lEclairage [14]. The parameters of the room were introduced in the software: the
reflection coefficients (walls and ceiling of 80% and 52% for the floor), the indoor objects and the
window type (double pane glazing with a visible transmittance of 90%). In Figure 6 is illustrated
only a single simulation that was realized for the date of 21 September at 12:00. Between the four
scenarios of WFR the indoor illuminance varies greatly from a mean value of 400 lux for WFR
15% to 797 lux for WFR 30%. We observe a clear increase of the natural lighting if increasing the
window area. This increase of visual comfort has also a repercussion on the electric energy
consumption which will decrease drastically if the WFR is higher.
15%
25%
20%
30%
Fig. 6 - Illuminance level for the WFR 15%, 20%, 25% and 30% - Simulation time 21 September at 12:00
27
4. Conclusions
The data obtained in this article were found from experimental measurements and from the
numerical analysis of the window to floor area ratio (WFR). In this research paper the indoor
comfort is studied experimentally during one day using professional equipment that allowed the
measurement of air quality, air temperature, humidity and illuminance level. The measurements
showed that the air temperature is comfortable for the intellectual activities even if the outdoor
air was around 11oC. The CO2 level was found to be much higher that the value proposed by the
standards with values of 2800 ppm. Despite the cloudy day, the indoor visual comfort is
achieved by both natural and artificial lighting. The glazing area is a key element in the design of
new classrooms, and that is the reason for studying different cases of WFR. If for the operative
temperature there is a slight difference between the WFR 15% and WFR 30% for the
heating/cooling consumption we have larger differences. A larger area of window signifies lower
energy consumption for the artificial lighting. The energy reduction can go up to 51% if using a
WFR of 30% compared to 15%. It can be concluded that a low WFR is not sufficient for a good
daylight while a higher WFR will increase the energy consumption. It is recommended that the
WFR should not overpass 25% or to be lower than 20%.
Acknowledgements
This work was supported by a grant of the Romanian National Authority for Scientific Research,
CNCS UEFISCDI, project number PN-II-RU-TE-2012-3-0108.
References
[1] A.C.K. Lai, K.W. Mui, L.T. Wong, L.Y. Law, An evaluation model for indoor environmental quality (IEQ)
acceptance in residential buildings, Energy and Buildings, Year 2009, Volume 41, Pages 930936.
[2] Tiberiu Catalina, Vlad Iordache, IEQ assessment on schools in the design stage, Building and
Environment, Volume 49, Pages 129-140, Year 2012
[3] Adelman H.S., Taylor L., Classroom climate, Encyclopedia of school psychology, Thousand Oaks, CA: Sage,
Year 2005.
[4] Grimsrud D., Bridges B., Schulte R., Continuous measurements of air quality parameters in schools, Building
Research and Information, Year 2006, Volume 34, Issue 5, Pages 447458.
[5] Bartlett K.H., Martinez M. and Bert j., Modeling of occupant-generated CO2 dynamics in naturally ventilated
classrooms, Journal of Occupational and Environmental Hygiene, Volume 1, Issue 3, Year 2004, Pages 139148.
[6] Faustman EM, Silbernagel SM, Fenske RA, Burbacher TM, Ponce RA., Mechanisms underlying children's
susceptibility to environmental toxicants, Environmental Health Perspectives, Year 2000, Volume 108, Issue 1,
Pages 13-21.
[7] World Health Organization Methods for monitoring indoor air quality in schools, Report of a meeting, Bonn,
Germany, 4-5 April 2011, Regional office for Europe.
[8] Daisey JM, Angell WJ, Apte MG., Indoor Air Quality, ventilation and health symptoms in schools: an analysis
of existing information, Indoor Air, Year 2003, Volume 13, Pages 53-64.
[9] J.Jalas, K.Karjalainen, P.Kimari., Indoor air and energy economy in school buildings, Proc. Of Healthy
Buildings,Year 2000, Volume 4, Pages 273-278.
[10] Becker R., Goldberger I., Paciuk M., Improving energy performance of school buildings while ensuring indoor air
quality ventilation, Building and Environment Volume 42, Issue 9, September 2007, Pages 32613276
[11] Mikola A., Voll. H, Koiv T., Rebane M., Indoor climate of classrooms with alternative ventilation systems,
GEMESED Proceedings of the 4th WSEAS international conference on Energy and development - environment
biomedicine, Year 2011.
[12] European Standard EN 15251, Indoor environmental input parameters for design and assessment of energy
performance of buildings addressing indoor air quality, thermal environment, lighting and acoustics, August 2007.
[13] ASHRAE. ANSI/ASHRAE Standard 55-2004 Thermal Environmental Conditions for Human Occupancy.
Atlanta : American Society of Heating, Refrigirating and Air-Conditioning Engineers, 2004. ISSN 1041-2336.
[14] CIE Technical Comittee 4.2, The Availability of Daylight, Technical Report No. NR, Comission
Internationale de lEclairage, Paris, 1975
28
INFLUENCE OF LONGITUDINAL VORTICES ON HEAT TRANSFER

FOR AIRFLOW PASSING THROUGH AN INNOVATIVE SOLAR
FACADE
Cristiana Verona CROITORU-Lecturer, PhD, Technical University of Civil Engineering Bucharest, Faculty of
Building Services, Politehnica University of Bucharest;
Florin BODE-Lecturer, PhD, Technical University of ClujNapoca;
Ilinca NASTASE-Associate Professor, PhD, Technical University of Civil Engineering Bucharest, Faculty of
Building Services, e-mail: cristianaveronacroitoru@gmail.com
Abstract: Renewable energy represents an attractive solution to fulfil two requirements: indoor
environmental quality and energy efficiency. Passive solar systems are easy to implement and
effective in areas with high solar potential. The Unglazed Transpired Solar Wall (UTSW) is made
of metal cladding with perforations, installed at several centimetres from a building wall, creating a
cavity. The air is forced to pass through this heated perforation, and thus a heat transfer between the
fluid and the metal takes place. Several measurements and CFD simulations were performed on an
innovative perforated solar wall. This study is a preliminary analysis approach on the importance of
the orifice shape of the perforated panel as a heat transfer parameter. The results found in literature
were compared with experimental and CFD results. A good agreement was found. Changing the
geometry of the perforations will increase on one hand the perforation perimeter and on the other
hand will generate complex fluid dynamics, resulting in a higher efficiency of heat recovery of
these devices.
Keywords: unglazed transpired solar wall; solar energy; energy efficiency; CFD modelling
1. Introduction
The new European Directives concerning energy performance of buildings imposes significant
reduction of the energy consumption. For this reason, the EU Members have adopted drastic
regulation in order to achieve high building performance. On the other hand, the indoor quality
has become an important parameter when conceiving residential or office buildings. The requests
of the occupants are more exigent and achieving the indoor comfort is one of the most important
challenges for civil engineers. Generally, the buildings sector consumes 35.3% from the total
energy demand. This energy demand is caused mainly by the HVAC (Heating Ventilating and
Air Conditioning) Systems. During the winter season in cold countries, the heat demand of the
building represents the highest percentage from the total amount of energy demand, while during
the summer, air treatment or ventilation is a major consumer of electrical energy. In this context,
the use of renewable energies is an attractive solution for fulfilling the two requirements: indoor
environmental quality and energy efficiency. Among renewable energies, the use of solar passive
systems are easy to implement and efficient from the accessibility point of view in the zones
with solar potential.
The Unglazed Transpired Solar Wall (UTSW) is made of metal cladding with perforations,
installed at several tens of centimetres from a building wall, thus creating a cavity through which
air circulates.
The schematic drawing of this type of solar collector is as illustrated in Fig. 1. The metal
cladding is heated by the solar radiation from the Sun and ventilation fans create negative
pressure in the air cavity, drawing in the solar heated air through the perforated panel. The air is
generally taken off the top of the wall (due to air temperature gradients in the cavity) ensuring
that all of the produced solar heat is collected, and then distributed in the building via the
ventilation system. In the summer conditions, the system can work only during the night for free
cooling ventilation, while during the day the air layer has an insulation role.
29
A literature survey led us to some interesting conclusions: (i) a consequent part of the heat
transfer between the air and the solar collector is occurring during its passage through the
perforation orifices; (ii) it is preferable to have a non-uniform flow on the back of the plate.
a)
b)
Fig.5 - a) Schematic drawing of an unglazed transpired wall, b) Innovative perforated panel developed at ULR [1]
On the other hand, passive mixing techniques applied to HVAC air diffusion terminal units have
been developed greatly during the past decade, since a collaborative research team from the
University of La Rochelle and UTCB dedicated numerous studies to these devices [2-11]. A new
research direction has been started at La Rochelle University (ULR) regarding the possibility of
using passive control for enhancing heat transfer in impinging jet flows [12]. All these studies
use a special geometry of nozzles, ailerons or orifices, which is called lobed geometry. An
example of such geometry is the lobed orifice. In Fig.1 b a perforated panel with lobed orifices
(cross shaped or 4-lobed orifices) is presented. For the same effective area (same equivalent
diameter) the perimeter of the lobed jet is much larger than the one of the circular orifice,
increasing the contact boundary between the air flow passing through the orifice and the orifice
thickness. Under low or moderate Reynolds numbers, such as the one characterizing the flows in
the UTSW, the analysis of the elementary lobed nozzle and orifice jets shows that the lobed
shape introduces a transverse shear in the lobe troughs [7, 13-15].
2. Methods
2.1. Experimental approach
The perforated panels were placed on a rectangular box with thermally insulated walls. The box
is connected through a circular pipe to an exhausting fan, forcing the ambient air to pass through
the perforated panel. After positioning each perforated cladding, the box is sealed with sanitary
silicone, in order not to have leaks which might perturb the tests.
At a distance of 30 cm of the cladding, four Metal Halide Flood Lights were placed, each
corresponding to a lightning level of 400 W, which are simulating the Sun radiation. The
aspiration fan creates the negative pressure necessary to force the air to pass through the
perforations, so the convective heat transfer takes place. Velocity, pressure and temperature
probes were placed in strategic points, either for controlling the conditions, or obtaining the
results of the measurements, as can be seen in the next figure. An acquisition data box was
employed to record the measurement values for a certain stabilisation period. The acquisition
time step is 60 seconds.
30
Perforated panel
Fan
Insulation
t1
t2
t3
t4
a)
b)
Fig.6 - a) Schematic drawing of the experimental facility b) Experimental set-up photograph: Radiation lamps,
exhaust pipe and the perforated panel
The cladding has a perforated surface of 0.49m2 (0.7m x 0.7m) and receives only a certain
percentage of the total radiative intensity emitted by the lamps. Four temperature probes were
placed in strategic points: t1- exit temperature; t2- ambiental temperature; t3- cladding surface
temperature; t4- black globe temperature. The velocity and pressure probes were used to evaluate
the airflow and pressure loss of the perforated sheet. The flow rate was evaluated using the
omnidirectional velocity probe from TSI, placed inside the exhaust pipe. For all the
measurements, we waited for the stabilisation time, each time three different readings of the
values being done.
The indoor temperature and relative humidity were permanently monitored. Several types of
perforated panels were tested: the baseline panel with round shape perforations and the
innovative panel with lobed cross-shaped perforations. The equivalent diameter De of both
geometries of orifices was 5mm. The porosity for each type of tested perforated panel is given by
the distance between two adjacent orifices, from centre to centre, of 13.5 mm (3C and 3R) and
19 mm (4C and 4R) for each type of perforation.
The perforated panels are positioned on a rectangular box with thermally insulated walls. The
box is attached through a circular pipe to an exhausting fan, forcing the air to pass through the
perforated panel, fact that conducted to the heating of the air.
The plate collects only a certain percentage of the total radiative intensity emitted by the lamps.
The radiation transmitted effectively between the source of light and the plate was considered to
be around 800 W/m2 (value in agreement with the experimental conditions from [16]).
2.1. Numerical approach
The numerical simulations by the CFD approach using a RANS (Reynolds Averaged Navier
Stokes) model were performed to study the airflow and heat transfer through the two types of
perforations for different values of the airflow.
31
Given some considerations of symmetry and in order to save numerical resources for a finer
mesh, the numerical study was performed for a smaller perforated cladding corresponding to a
part of the experimental panel. The model comprises 25 perforations (Fig. 3), with the same
space between two adjacent centers of orifices, as in the experimental case. The metal cladding
has a size of 10 cm by 10 cm positioned at 15 cm from the exhaust surface. This approach will
test the capabilities of the CFD models to reproduce from the global point of view our previous
experimental results [17] in order to investigate in a next step which the influence of the flow
dynamics is on the heat transfer enhancement.
Fig. 7 - Section through the used grid and detail of the cross perforated cladding
The boundary conditions on the perforated cladding considered an imposed thermal flux such as in
the experimental case. The rest of the walls are considered with 0 W heat-fluxes. The air is aspired
through the perforations as in the experimental case. A value of turbulence intensity of 9.8%,
calculated with empirical relation proposed by [18], was imposed for the inlet boundary condition.
The accuracy of a CFD simulation depends, in a high percentage, on the way of replicating the
geometry that defines the calculation domain and the heat sources, with specific boundary
conditions. The final computational domain comprises 4.5 million hybrid cells: both tetrahedral
and hexahedral cells for a better characterization of the flow. Inside the orifices a first layer of
0.2 mm was applied with a growth factor of 1.15. Outside the orifices, the mesh on the plate has
a first layer of 5 mm, with a growth factor of 1.15. The viscous model was chosen to be k-omega
SST in agreement to previous studies performed on the lobed perforations[19].
3. Results
The studies performed for solar walls systems showed good results in terms of energy efficiency. In
this context the use of solar passive systems is encouraged by national regulations as they can have a
significant contribution to achieve high performances and to save energy for winter heating and for
summer cooling. Table 1 summarizes some of the case studies available in the literature.
A quick survey allows us to be aware of the huge possibilities of such devices in energy recovery.
For instance, the CFD study of Arulanadam et al. [20] concludes that not only metal cladding could
be used for the perforated absorber but even low conductivity materials can lead to acceptable
thermal efficiency of the system, for low porosity of the transpired plate absorbers and for low
velocity flow situations. But studies such as the ones of Van Decker et al. [21], Gunnewieck et al.
[22, 23] are very interesting from our point of view, given the information related on the direct
possibilities of improvement of these devices. The early numerical study of Gunnewieck et al. [22]
highlights the importance of a non-uniform flow and of a low velocity on the efficiency of unglazed
transpired solar air heaters of large area. Van Decker et al. [21] show that in no-wind conditions,
about 62% of the ultimate temperature rise of the air is predicted to occur on front-of plate, 28% in
the hole and 10% on the back of the plate.
32
Cordeau and Barrington [24] in their study of an UTSW, used for bringing fresh air in a broiler barn,
reveal that the efficiency of the solar air pre-heaters reached 65% for wind velocities under 2 m/s,
but dropped below 25% for wind velocities exceeding 7 m/s, with an annual return on investment of
4.7%. Different other case studies of UTSW [16, 22-32] pointed out an energy efficiency of the
system used from 52% to 68%, being an important benefit in terms of fossil energy consumptions
savings.
Table 1
Current studies on UTSW
Reference
[33]
[33], [34]
[33]
[35]
[35]
[36]
Present
study
Collector area and type

1877 m2; vertical wall; 2%
porosity; 1% canopy;
420 m2 gross; 2%porosity; 1%
canopy
27.9 m2; 2% porosity
335 m2; corrugated dark brown
aluminum
Solar wall panel area=1.1664 m2;
PV cells covered 24% of entire
surface
2 m; Transpired solar collector
1 m2 ; Transpired solar collector;
0.6 % - 10% porosity; black
aluminum sheet
Airflow
rate
[m3/h/m2]
125
12.5 C
57%
917 kWh/m2/year
72
13 C
52%
754 kWh/m2/year
N/A
N/A
N/A
N/A
63-68%
N/A
N/A
195 700 kWh/m2
100
N/A
Thermal
efficiency 48%
Combined
efficiency 51%
500-1000
kWh/m2/year
From
which
electricity
50-100 kWh/m2/year
117
10-150
[m3/h/m2]
13.2C
9C-30C
60-70%
for
airflows larger
than
50
[m3/h/m2]
N/A
Temperat
ure rise
Efficiency
Energy saving
In the present study, the heat transferred from the plate to the air (P) was quantified by the air
temperature rise, using:
P=mair* cp *(Tpipe-Tamb)
(1)
where mair is the mass flow rate.

Four cases were studied in comparison with a standard configuration of a commercial perforated
panel for UTSW systems. The studied cases are: 3R - Round orifices with S= 13.5 mm, 4R - Round
orifices with S= 19 mm, 3C Cross shaped orifices with S=13.5 mm and 4C Cross shaped orifices
with S=19 mm.
The efficiency of the panel was defined as:
P
I T Apl
(2)
where P is the heat transferred from the plate to the air, IT is the irradiation provided by the lamps to
the plate level and A is the surface of the plate of 1 m2.
In Fig. 4 we represented the evolution of the thermal efficiency for the four cases investigated, in
comparison with the data obtained by using a commercial UTSW and the two models proposed by
Belusko et al. et Shukla et al. for UTSW without wind and with circular perforations with the same
equivalent diameter as in our case [30, 37]. For both perforation rates studied in the present
experimental campaign we can see an advantage of the innovative perforated plate with lobed
orifices compared to the baseline round orifices panels. They present also a clear advantage for high
flow rates when compared to the analytical models of Belusko et al. et Shukla et al. for panels with
circular perforations.
33
70
70
60
Efficiency(%)
Efficiency(%)
60
3R
50
3C
40
CommercialUTSW
3RDymondetal.1999
30
4R
4C
50
CommercialUTSW
40
4RDymondetal.1998
4RBeluskoetal2008
30
3RBeluskoetal2009
20
20
10
10
0.00
50.00
100.00
150.00
200.00
0.00
50.00
a)
Q(m3/h/m)
100.00
150.00
200.00
Q(m3/h/m)
b)
Fig. 8 - UTSW efficiency for different volumetric flow rates: a) S=13.5, b) S=19 mm
If we compare the heat transfer that occurs, we can observe that the cross shape performs better, for
both cases: 4C and 3C.
300
P(Watts)
250
200
150
RoundOrifices3R
CrossOrifices3C
Roundorifices4R
100
Crossorifices4C
Referenceperforatedpannelfromcommercialsystem
50
0
20
40
60
80
100
120
140
Q(m3/h)
Fig. 9 - UTSW efficiency for different volumetric flow rates: a) S=13.5, b) S=19 mm
We wanted further to test a larger domain of the airflow employed in the experimental setup for the
metal cladding with perforation of 13.5 mm (3C and 3R). We evaluated the heat transfer for airflows
ranging from 30m3/h up to 270 m3/h. We can observe that after a stagnant zone of heat transferred,
for values of airflow of 100 m3/h we obtain a significant increase in thermal power. These findings
will be treated in a further study.
550
500
450
400
P(Watts)
350
300
3C
250
3R
200
150
100
50
20
30
80
130
Q(m3/h)
180
230
280
Fig. 10 - Heat transfer for a larger airflow domain: 3C and 4R comparison
Let us take a look to the CFD results compared to the experimental data. In Fig. 6 and 7 we present
typical velocity and temperature fields in a plane passing through one of the perforation rows
adjacent to the median plane, for both geometries: cross and round shape. We can observe an
increase in temperature between inlet and outlet, intensified in the case of the cross perforation.
34
a)
b)
Fig. 11 - CFD fields for the round perforations at a flow rate of 96 m3/h: a) velocity, b) temperature
a)
b)
3
Fig. 12 - CFD fields for the cross-shaped perforations at a flow rate of 101 m /h: a) velocity, b) temperature
In Fig. 8 we superposed experimental and numerical data for the temperature differences obtained for
the two types of perforations. While in the case of the round shape perforation the temperature
difference between the ambient temperature and the temperature of the airflow aspirated through the
perforated plate are little underestimated, in the case of the lobed perforation they are rather
overestimated. We notice however the similarity between the experimental data of Leon et al. [16]
and our numerical data.
20
18
Experimental
16
DeltaT(K)
12
DeltaT(K)
CFD
14
Leonetal.(2007)
10
8
6
4
2
0
0
50
100
150
200
250
20
18
16
14
12
10
8
6
4
2
0
Experimental
CFD
Q(m3/h)
50
100
150
200
250
Q(m3/h)
Fig. 13 - Temperature difference as a function of the airflow: a) circular perforations b) cross shaped perforations
In Fig. 9 thermal efficiencies for the two UTSW are given from experimental and numerical cases, in
comparison with the data obtained by using a commercial UTSW and the two models proposed by
Belusko et al. et Shukla et al. for UTSW without wind and with circular perforations with the same
equivalent diameter as in our case [30, 37]. For both perforation rates studied in the present
experimental campaign we can see an advantage of the innovative perforated plate with lobed
orifices compared to the baseline round orifice panels.
35
100
Efficiency(%)
90
80
70
60
50
40
30
20
10
0
50
Q(m3/h/m)
100
150
200
Round CFD
Cross CFD
Round Experimental
Cross Experimental
Dymondetal.(1998)
Beluskoetal.(2008)
Fig. 14 - UTSW efficiency for different volumetric flow rates; Comparison with literature
They also present a clear advantage for high flow rates when compared to the analytical models of
Belusko et al. and Shukla et al. for panels with circular perforations.
4. Conclusions
The study evaluated the energy efficiency of several types of unglazed transpired solar collector
(UTSW) by experimental and numerical means. The physical model used shows good results in
agreement with literature. In addition, the comparison of a conventional UTSW with a new
geometry with innovative perforation leads to interesting results, with over 15% increase in
thermal efficiency since the literature shows a lack of the geometry study for the perforations.
These effects still need to continue the investigation. The CFD study on the unglazed transpired
solar collector (UTSW) which is equipped with cross shape perforations shows that the
experimental conclusions can be also found by numerical means. Because such geometries
require very fine meshes, a scaled model of the experimental would be the answer to numerical
modelling for such case. The results showed very good agreement with the experimental study,
fact that validated our model. The efficiencies calculated proved the advantage of cross-shaped
models in comparison to classical ones. More of that, the comparison of a classical UTSW with a
new one with innovative perforation geometries leads to interesting results, with more than 15%
increase in thermal efficiency for volumetric flow rates higher than 100 m3/h. Further studies
regarding the best configuration still have to be conducted, including complementary
measurement techniques.
Acknowledgements: This work was supported by the grants of the Romanian National
Authority for Scientific Research, CNCS UEFISCDI, project number PN-II-RU-PD-2012-30144, PN-II-ID-PCE-2011-3-0835.
References
[1] Meslem, A., I. Nastase, and F. Allard, Passive mixing control for innovative air diffusion terminal devices for
buildings. Building and Environment, 2010. 45: p. 2679-2688.
[2] Meslem, A., I. Nastase, and F. Allard, Passive mixing control for innovative air diffusion terminal devices for
buildings. Building and Environment, 2010. 45 ( 2679-2688).
[3] Meslem, A., I. Nastase, and K. Abed-Meraim, Experimental investigation of a lobed jet flow mixing
performance. Journal of Engineering Physics and Thermophysics, 2007. 81(1).
[4] Meslem, A., M. El-Hassan, and I. Nastase, Analysis of jet entrainment mechanism in the transitional regime by
time-resolved PIV. Journal of Visualization, 2010. online first: p. 1-12.
[5] Meslem, A., et al., A comparison of three turbulence models for the prediction of parallel lobed jets in
perforated panel optimization. Building and Environment, 2011. 46(11): p. 2203-2219.
[6] Nastase, I., et al., Lobed grilles for high mixing ventilation - An experimental analysis in a full scale model
room. Building and Environment, 2011. 46( 3): p. 547-555.
[7] Nastase, I., A. Meslem, and P. Gervais, Primary and secondary vortical structures contribution in the
entrainement of low Reynolds number jet flows. Experiments in Fluids, 2008. 44(6): p. 1027-1033.
36
[8] Nastase, I., A. Meslem, and M. El Hassan, Image processing analysis of vortex dynamics of lobed jets from
three-dimensional diffusers. Fluid Dynamics Research, 2011. 43(6).
[9] Nastase, I. and A. Meslem, Vortex Dynamics and mass entrainement in turbulent lobed jets with and without
lobe deflection angles. Experiments in Fluids, 2010. 48(4): p. 693-714.
[10] Nastase, I. and A. Meslem, Vortex dynamics and entrainment mechanisms in low Reynolds orifice jets. Journal
of Visualization, 2008. 11(4).
[11] Nastase, I. and A. Meslem. Lobed jets for improving air diffusion performance in buildings. in The 29th AIVC
Conference. 2008. Kyoto, Japon.
[12] Kristiawan, M., et al., Wall shear rates and mass transfer in impinging jets: Comparison of circular convergent
and cross-shaped orifice nozzles. International Journal of Heat and Mass Transfer, 2012. 55(13): p. 282-293.
[13] Nastase, I. and A. Meslem, Vortex Dynamics and Entrainment Mechanisms in Low Reynolds Orifice Jets.
Journal of Visualisation, 2008. 11(4): p. 309-318.
[14] Nastase, I. and A. Meslem, Vortex dynamics and mass entrainment in turbulent lobed jets with and without lobe
deflection angles. Experiments in Fluids, 2010. 48(4): p. 693-714.
[15] El-Hassan, M. and A. Meslem, Time-resolved stereoscopic PIV investigation of the entrainement in the nearfield of circular and daisy-shaped orifice jets. Physics of Fluids, 2010. 22.
[16] Leon, M.A. and S. Kumar, Mathematical modeling and thermal performance analysis of unglazed transpired
solar collectors. Solar Energy, 2007. 81(1): p. 62-75.
[17] Croitoru, C., Meslem, A., Nastase, I., Martin, O. Innovative Solution of Unglazed Transpired Plate Solar Air
Heater. in Clima 2013. 2013. Prague.
[18] Jaramillo, J.E., et al., Numerical study of plane and round impinging jets using RANS models. Numer. Heat
Transfer Part B 2008. 54: p. 213-237.
[19] Florin BODE, I.N., Cristiana CROITORU, RANS models comparison for a cross-shaped jet flow with straight
lobes. Mathematical Modelling in Civil Engineering, 2012. 8(4): p. 14-20.
[20] Arulanandam, S.J., K.G.T. Hollands, and E. Brundrett, A CFD heat transfer analysis of the transpired solar
collector under no-wind conditions. Solar Energy, 1999. 67(13): p. 93-100.
[21] Van Decker, G.W.E., K.G.T. Hollands, and A.P. Brunger, Heat-exchange relations for unglazed transpired solar
collectors with circular holes on a square or triangular pitch. Solar Energy, 2001. 71(1): p. 33-45.
[22] Gunnewiek, L.H., E. Brundrett, and K.G.T. Hollands, Flow distribution in unglazed transpired plate solar air
heaters of large area. Solar Energy, 1996. 58(46): p. 227-237.
[23] Gunnewiek, L.H., K.G.T. Hollands, and E. Brundrett, Effect of wind on flow distribution in unglazed
transpired-plate collectors. Solar Energy, 2002. 72(4): p. 317-325.
[24] Cordeau, S. and S. Barrington, Performance of unglazed solar ventilation air pre-heaters for broiler barns. Solar
Energy, 2011. 85(7): p. 1418-1429.
[25] Hollick, J.C., Unglazed solar wall air heaters. Renewable Energy, 1994. 5(14): p. 415-421.
[26] Hollick, J.C., Solar cogeneration panels. Renewable Energy, 1998. 15(14): p. 195-200.
[27] Konttinen, P., T. Salo, and P.D. Lund, Degradation of unglazed rough graphite-aluminium solar absorber
surfaces in simulated acid and neutral rain. Solar Energy, 2005. 78(1): p. 41-48.
[28] Sopian, K., et al., Performance of a non-metallic unglazed solar water heater with integrated storage system.
Renewable Energy, 2004. 29(9): p. 1421-1430.
[29] Nblek, B., et al., Performance of an unglazed textile-plastic solar absorber. Renewable Energy, 1999. 16(14):
p. 635-638.
[30] Belusko, M., W. Saman, and F. Bruno, Performance of jet impingement in unglazed air collectors. Solar
Energy, 2008. 82(5): p. 389-398.
[31] Molineaux, B., B. Lachal, and O. Guisan, Thermal analysis of five unglazed solar collector systems for the
heating of outdoor swimming pools. Solar Energy, 1994. 53(1): p. 27-32.
[32] Kumar, S. and S.C. Mullick, Wind heat transfer coefficient in solar collectors in outdoor conditions. Solar
Energy, 2010. 84(6): p. 956-963.
[33] Cali, A., Kutscher, C.F., Dymond, C.S., Pfluger, R., Hollick, J., Kokko, J., et al. , A report of Task 14 Air
Systems Working Group: low cost high performance solar airheating systems using perforated absorbers, in
1999, International Energy Agency (IEA): Washington.
[34] Hollick, J.C., Unglazed solar wall air heaters. Renewable Energy, 1994. 5: p. 415-21.
[35] Hollick, J.C., Worlds largest and tallest solar recladding. Renewable Energy, 1996. 9: p. 703-7.
[36] Chan, H.Y., S. Riffat, and J. Zhu, Experimental performance of unglazed transpired solar collector for air
heating, in World Renewable Energy Congress. 2011: Sweden.
[37] Shukla, A., et al., A state of art review on the performance of transpired solar collector. Renewable and
Sustainable Energy Reviews, 2012. 16(6): p. 3975-3985.
37
DESIGN PROCEDURE FOR SIDE WALLS OF SOCKET FOUNDATIONS

Ionu DAMIAN - Assistant, Technical University of Civil Engineering, Faculty of Civil, Industrial and Agricultural
Structures, Department of Reinforced Concrete, e-mail: ionutzdamian@gmail.com
Abstract: Single storey structures having simple structural systems, jointed roof on cantilever
columns are widely used nowadays for commercial buildings. To minimize the execution time and
construction effort, a popular solution is to install precast columns on the so called socket
foundations. The column is a linear element characterized by flexural behavior, the design of which
does not produce difficulties. However, the foundation is made of several short elements, the
behavior of which may be difficult to assess. One web element is the side wall of the socket. This
element behaves like a short cantilever with the load suspended at the bottom. The new version of
the Romanian code for shallow foundations, NP 112/2012 proposes a strut and tie model for the
design of the side wall. Anyway, the model is not fully described and the strut capacity is not
verified against the results of more refined analyses.
The goal of the paper is to complete the strut and tie model and to present a simple design relation
for estimating the capacity of the struts. The method is calibrated against the predicted monotonic
behavior calculated using the Modified Compression Field Theory (MCFT), a formulation
adequate for the analysis of squat elements. The article also includes a design example using the
proposed procedure.
Keywords: socket foundation, side wall, Modified Compression Field Theory (MCFT), strut and tie
1. Introduction
Single storey structures having simple structural systems, jointed roof on top of cantilever
columns are frequently used for commercial buildings. The current state of practice is to use
precast elements for both columns and isolated footings. The design of the columns is not
difficult, because they are linear members characterized by flexural behavior. However, the
foundation is a complex element, the behavior of which is made difficult to assess, especially
when it is subjected to significant bending moment.
The loads are transmitted from the column to the foundation within a back stay mechanism
through a system of side walls called socket. In fact, this foundation system is called socket
foundation. The front wall is subjected to a bending moment produced by the pressure of the
column (fig. 1.b). The out of plane shear force on the front wall of the socket is transmitted to the
side wall as in plane shear force (fig. 1.b).
It can be noticed that the front wall of the socket behaves like a plate, while the lateral walls
behave like membranes. There are several design models for estimating the forces on the walls
of the socket. The mostly used model is the one proposed by Leonhardt and Molling [1], which
was also adopted in the Romanian code NP 112/2006 [2] for the design of shallow foundations.
According to that code, the lever arm of the horizontal force is 80% of the socket height (fig.
1.a), which takes to the following lateral force:
VEd
M Rm
Vas
0.8 H s
(1)
where:
M Rm is the moment capacity of the column, calculated with average material strengths;
Vas is the value of the shear force in the column associated with moment capacity M Rm ;
H s is the socket height.
38
Fig. 1 (a) Transmission of the forces from the column to the socket by back stay mechanism ; (b) Bending moment
forces on the front wall and, respectively, shear forces on the side walls of the socket.
The NP 112/2006 [2] dealt with the design of the front wall for out of plane bending moments.
However, the new version of the code, NP 112/2012 [3] eliminates this rude design because the
contact forces are applied close to the side walls and the forces are carried directly by means of a
strut mechanism (fig. 1.b).
In the NP 112/2006 [2], the design of the side walls of the socket to in plane forces depends on
the aspect ratio of the wall. Strut capacity is limited by the value of the allowable shear stress in
the section of the wall, according to the relation:
Ed
0.5VEd
2 fctd
tw Ls
(2)
where:
t w is the thickness of the side wall;
Ls is the side wall depth;
fctd is the design tensile strength of the concrete.
The NP112/2012 [3] provides a strut and tie mechanism to design the side wall of the socket (fig.
2.a and c). The axial force in the concrete strut is calculated with the formula:
Cw.Ed
0.5VEd
cos
(3)
where is the strut angle.

The thickness of the wall results from the criterion:
w.Ed
Cw.Ed
Rd .max
bw tw
(4)
where bw is the width of the strut (fig. 2.c).

The value of the maximum compression stress in the strut is calculated with the formula:
Rd .max 0.75 fcd

(5)
39
where:
f cd is the design compression strength;
is a strength reduction factor which takes into account for the bi-axial state of stresses. It is
calculated with the formula:
f ck
250
where fck is the characteristic compression strength of the concrete in MPa.
0.6 1
(6)
Although the design model used in the new version of the code is more refined when compared
to the one used in the old version, it has three main shortcomings:
1) the angle of the strut is not defined;
2) the formula used for the calculation of the width of the strut is odd-to say the leastdepending on the anchorage length of the reinforcement;
3) the strut resistance is not calibrated for the special case of the socket foundations. It is
taken from EN 1992-1-1 [4] as the resistance of the traction-compression nodes used for
the strut and tie models.
Therefore, the objective of the article is to fully substantiate the strut and tie model used for the
design of the side wall of the socket. This requires the following steps:
1) proposing a reasonable based angle for the strut;
2) proposing a reasonable width of the strut;
3) calibrating the resistance of the strut for the proposed strut and tie model.
Additionally, the resistance of the strut will also be defined by the mean shear stress on the
section, as was done in the old version of the code.
2. The Strut and Tie Model
The geometry of the strut is determined from the model presented in fig. 2.a. The angle of the
strut is calculated with the formula:
tan
Hs
Ls tw
(7)
For the width of the strut, the following formula results from geometrical consideration in
figure 2.b:
bw tw sin 0.4 H s cos
(8)
The only parameter that remains to be found is the strength of the strut. This is not an easy task,
taking into account that the side wall of the socket is a squat element. Static nonlinear analysis
will be used for the investigation of the concrete strut resistance. However, the nonlinearity
should capture the interaction between shear force, bending moment and axial force in the
membrane element. A simple, yet accurate enough analysis method that has been widely used for
the investigation of squat elements is the Modified Compression Field Theory (MCFT) proposed
by Vecchio and Collins [5] in 1986. This method will be used in this investigation as the main
analysis method.
40
Fig. 2 (a) Strut and tie model; (b) Geometrical consideration for calculating the width of the strut
(c) Strut and tie scheme
The steps in the calibration process are:

1) Propose a set of column sections, column lengths, target longitudinal reinforcement ratio and
the value of normalized axial force;
2) Propose a set of Hs/Ls ratios;
3) For each column dimensions proposed in (1) find a set of approximative Ls values based on
an approximate socket thickness (Ls=b+30...50cm, depending on column dimension, b);
4) For each Ls value determined at (3), calculate the socket height based on the ratios proposed in (2);
5) For each pair (Hs, Ls), design the reinforcement of the side wall of the socket using code
requirements. Design relations are presented in chapter 5. The horizontal tie reinforcement is
overdesigned using an overstrength factor of two in order to prohibit its yielding.
6) Propose wall thickness. For the initial guess, propose a thickness that will produce concrete
crushing.
7) For the proposed reinforcement and wall thickness, run static nonlinear analysis using the
MCFT method;
8) Determine the strength of the wall, FR and the collapse mechanism;
9) If the mechanism is longitudinal steel yielding, perform the calculations backwards to
determine strut resistance; otherwise return to (6) and propose a higher value of the
thickness. Combining relations (3) and (4), the following formula is used to calculate strut
resistance:
FR
R
cos bw tw
(9)
10) After running all analyses, perform statistics of the data and calibrate the strut resistance.
The proposed calibration method requires lots of analyses to have enough results to perform
credible statistics.
There are two main reasons why available traditional software is not adequate for this analysis:
i. there are only a few programs that perform MCFT analyses;
ii. the programs are not flexible, namely a new analysis model has to be manually
constructed for each design solution, which makes the study techy.
41
Therefore, a handmade finite element program that performs static nonlinear analyses using
MCFT has been developed in MATLAB, a well-known numerical analysis platform. It has the
advantage that constructs automatically a new side wall model based on input data. This program
is presented in the next section.
For the static nonlinear analysis, the side wall is isolated from the rest of the socket. The wall is
subjected to a horizontal force pattern like the one presented in fig. 3, which models code contact
force distribution (the hatch in fig. 1.a.). The horizontal forces are increased until the strength of
the wall, FR, is reached. Mean material strengths are used in the analyses.
Fig. 3 Horizontal force pattern of the finite element analysis model
3. Computer Program
3.1. MCFT
MCFT is a theory elaborated by Vecchio and Collins in the mid 80s to evaluate shear resistance
of square reinforced concrete panels. The popularity of the theory originates from its ability to
solve a problem that has puzzled the engineering society for decades: the interaction between
shear and bending, especially for squat elements.
MCFT is a smeared crack theory specific for reinforced concrete elements which considers that
cracks direction is normal to the direction of the principal tension strain (fig. 4). This basic
assumption leads to the rotation of the cracks during the analysis; therefore it is called a "rotating
angle theory". Reinforcement slippage is eliminated as an assumption.
Fig. 4 (a) Principal concrete stress; (b) Mohr circle of strains; (c) Mohr circle of concrete stresses
42
The theory uses three types of equations which are presented next.
i) Equilibrium equations. For an orthogonally reinforced panel, the following equations result
(fig. 5):
Fig. 5 Stresses in orthogonally reinforced panels
f x sx f sx fcx
f y sy f sy f cy
xy cxy
(10)
where:
sx , sy are the reinforcement ratios in the two orthogonal directions;

f sx , f sy are the steel stress in the two orthogonal directions;
fcx , fcy are the concrete stresses in the two orthogonal directions;
f x , f y are the total stresses in the two orthogonal directions;
xy , cxy are the total shear stress and, respectively, the shear stress in the concrete.
ii) Geometric conditions. From the Cartesian strains, the principal strains (fig. 4.b) are calculated
with the formula:
1,2
x y
2
x y xy

2 2
(11)
where:
x , y are the strains in the two orthogonal directions;

xy is the shear strain.
The angle of the principal strain direction is calculated with the formula:
1
2
xy
x y
arctan
(12)
Considering that 1 2 , the following relation should be true:
sign tan sign xy
(13)
43
iii) Material laws.

For the reinforcement, an elastic-perfectly plastic relation is used, as follows (fig. 6.a):
Es s , if s ys
fs
f ys , if s ys
(14)
where:
Es is the Youngs modulus of the reinforcing steel;
s is the steel strain;
ys is the yield strain of the reinforcement;

f ys is the yield stress of the steel.
For the concrete subjected to compression efforts, Hognestad parabola is used (fig. 6.c):
2
f c 2 f c 2.max 2 2 2
c0 c0
(15)
where:
c0 is the concrete strain at maximum stress, considered positive in relation (15);

f c 2.max is maximum concrete strength for tension-compression loading, which depends on
the orthogonal strain (fig. 6.d). It is calculated with the formula:
fc 2.max
fc
0.8 0.34 1
c0
fc
(16)
Fig. 6 Material laws for: (a) reinforcing steel; (b) concrete in tension; (c) concrete in compression. (d)
Compression strength reduction
44
The dependence of the concrete strength on the principal tension strain was noted by Vecchio
and Collins [5] when performing some tests on a series of square panels. This reduction of the
concrete strength actually gave the name of the theory: "compression field".
The cracking criterion in terms of tension stress is fulfilled when the cracking strain is exceeded.
After cracking, the concrete is considered to be capable of taking tension stresses due to the
tension stiffening effect. The overall behavior is described by the curve (fig. 6.b):
Ec 1, if 1 cr
f ct
fc1
, if 1 cr
1 c
t 1
(17)
where:
Ec is the initial Youngs modulus for concrete;
fct is the concrete tension strength;
cr is the concrete cracking strain;

ct is a parameter that takes into account tension stiffening effects after cracking.
Several values were proposed for this parameter. For example, Vecchio and Collins [5] proposed
a value of 200 based on their experimental results; Collins and Mitchel [6] proposed a value of
500 for this parameter based on the results on another set of panels, while Benz [7] proposed a
more rational formula that depends on the reinforcement ratios in the two orthogonal directions.
A special feature of MCFT is the crack check. The continuity of element efforts between section
1-1 in between cracks, and section 2-2 in the crack itself in fig. 7.a requires the fulfillment of two
equilibrium relations.
Fig. 7 (a) Cracked reinforced concrete panel; (b) Average stress in between cracks and (c) Stress in the crack
The two equilibrium relations can be written either in the directions normal and parallel to the
crack, or in the two Cartesian directions. The equilibrium equations in the normal and parallel
crack directions are (fig. 7.b and c):
fc1 sx f sxcr f sx cos2 sy f sycr f sy sin 2
ci sx f sxcr f sx sin cos sy f sycr f sy sin cos
(18)
where:
f sxcr , f sycr are the reinforcement stresses in the crack in the two orthogonal directions;
ci is the shear stress in the crack.

45
The continuity relations in the two Cartesian directions are (fig. 7.b and c):
sx f sxcr f sx fc1 ci tan

sx f sxcr f sx fc1 ci ctg
(19)
The steel stress in the crack cannot exceed the yielding stress; therefore the principal concrete
stress is limited to the value:
f c1 sx f ys f sx cos 2 sy f ys f sy sin 2
(20)
If the principal concrete stress calculated with relation (17) does not fulfill relation (20), it is
assigned to the right part of inequality (20). The shear stress in the crack is calculated with the
second equation in (18) and the reinforcement stresses in the crack are calculated with relations
(19). In this case, the curve for tensioned concrete has the branch (b) in fig. 6.b, otherwise it
continues with branch (a) in the same figure.
It should be noted that there are only two independent equilibrium equations and three unknowns
to be determined in the crack. Therefore, excepting the aforementioned case, the stresses in the
crack are determined based on the principle of the minimum shear stresses, considering
reinforcement yielding. From the two equations in relation (19) results:
1ci
sx f ys f sx f c1
tan
ci2 f c1 sy f ys f sy tan
There are several possible solutions.
(21)
1) If sx f ys f sx f c1 sy f ys f sy , it results that the shear stress in the crack is

positive and one of the reinforcements yields. It is calculated with the relation:
ci min 1ci , ci2
(22)
2) If sx f ys f sx fc1 and sy f ys f sy fc1 , it results that there is no shear stress in the

crack. In this case, no reinforcement reaches yielding.
3) If sy f ys f sy f c1 sx f ys f sx , it results that the shear stress in the crack is

negative and one of the reinforcements yields. It is calculated with the relation:
ci max 1ci , ci2
(23)
4) If sx f ys f sx fc1 and sy f ys f sy fc1 , it results that there is no shear stress in the

crack. Furthermore, there is no reinforcement yielding.
With the known shear stress in the crack, relations (19) are then used for the calculation of the
reinforcement stresses.
It should be mentioned that the implemented MCFT version does not limit the shear stress in the
crack as in the original model because of two reasons:
i.
46
It could be demonstrated that decreasing the shear stress in the crack results in increasing
the principal concrete stress, which is not possible. To be more precise, in the original
MCFT formulation it was possible to propose other values of the strains to result in a
lower value of the shear stress in the crack because only one panel was analyzed.
However, the finite analysis works with a large number of elements and the
aforementioned procedure may not always have a solution.
ii.
The shear stress in the crack is generally small if compared to the allowable shear stress,
so neglecting it has no impact on the results.
For the calculation of the crack width, the method proposes the following relation:
w sr 1
(24)
where sr is the crack spacing, which depends on the crack spacings srx and sry in the two
orthogonal directions, generally calculated according to the code provisions.
The crack spacing is calculated with the formula:
sr
cos sin
srx
sry
(25)
It should be specified that for the present implementation, crack spacing has no influence on the
results because shear crack check is not performed.
3.2. MCFT Implementation
Although in its original form, the MCFT was used for hand computations, Vecchio extended its
application to finite element programs. The implementation in this paper is similar to the one
presented by Vecchio [8]. This chapter will present the principal issues of the implementation.
The program uses rectangular isoperimetric membrane elements with four nodes. In this case, the
interpolation function for displacements has a linear variation depending on the displacements of
the corner nodes.
For the simplicity of the formulation, element Cartesian strains are evaluated in the center of the
element with the formula:
xy Bc del
(26)
where:
xy is the Cartesian strain vector;
Bc is the strain interpolation matrix evaluated at the center of the element;

del is the vector of element nodal displacements.
This option has two main advantages:
i.
It avoids the so called shear locking phenomenon, which produces stiffer elements in
shear loading;
ii.
It allows the use of closed forms for the generation of the stiffness matrix, avoiding
computationally expansive numerical integration. It must be specified that at least four
Gauss integration points have to be used for full integration if the aforementioned method
is not used. The computational tests performed for the development of the program
showed that one integration point produces stiff elements that make convergence difficult
after concrete cracking.
Another issue of the numerical implementation of the method is the solver. There are two types
of solvers that are currently implemented in finite element analyses software. The first type, and
the most popular, is the solvers based on the tangent stiffness matrix. The other type is based on
secant stiffness matrix, which was used in Vecchio [8].
47
The program has incorporated both types of solvers. However, the secant stiffness matrix solver
is superior to the tangent stiffness matrix solver, performing significantly faster analyses. The
key issue on computational time is the cracking region. Tangent stiffness solver crosses with
difficulty this region, although powerful numerical methods such as Criestfields arc length [9],
Bathes constant work increment [10] or Powells constant displacement increment [11] methods
are implemented. The only viable explanation is that concrete cracking produces successive load
reductions which are difficult to handle. Furthermore, it should be observed that the secant
stiffness approach has the advantage of a symmetric stiffness matrix, unlike tangent stiffness
matrix solver, which saves computational time because specific linear equation solvers are used.
This is the reason why the secant stiffness approach is presented in this chapter.
The secant stiffness matrix of the element is computed from the material stiffness matrix. The
material stiffness matrix is the sum between concrete stiffness matrix and reinforcement stiffness
matrix:
D Dc Ds
(27)
The material stiffness matrix of the concrete in principal strain axes has the following form:
Ec1
0
0
Dc,12 0 Ec 2
0
0
0 Gc,12
(28)
The diagonal elements represent the secant material elasticity modulus (fig. 6 b and c) calculated
based on equations (15), (16) and (17):
f
Ec1 c1
f
Ec 2 c 2
Gc,12
Ec1 Ec 2
Ec1 Ec 2
(29)
Although the secant shear modulus is small after cracking, it ensures the stability of the solution
in the first loading steps when shear stresses cannot be taken by the strut mechanism since the
struts are almost vertical.
The material stiffness matrix of the concrete in Cartesian coordinates is evaluated with the
formula:
Dc TT Dc,12 T
(30)
The strain rotational matrix is calculated as:
cos 2
sin 2
sin cos
sin cos
T
sin 2
cos 2
2 sin cos 2 sin cos cos 2 sin 2
(31)
For orthogonally reinforcement arrangement, the material stiffness matrix of the reinforcement is
calculated with:
sx Esx
Ds 0
0
48
0
sy Esy
0
0
0
(32)
incr=1
Assembly initial secant stiffness matrix, Ks
Propose the new value of external force vector Fext

ite=1
Solve for the equation system Fext=Ks d, resulting the displacement vector, d
incr=incr+1
iel=1
Evaluate strains in the center of the element, xy=Bc del

Calculate reinforcement efforts fs and material stiffness matrix Ds
Calculate principal strains, 1 and 2
Calculate principal concrete stresses fc1 and fc2 and material stiffness matrix Dc,12
Calculate concrete material stiffness matrix in Cartesian system Dc=TT Dc,12 T
Calculate total material stiffness matrix in Cartesian system D=Dc+Ds
Calculate element stiffness matrix, Kel
iel=iel+1
ite=ite+1
Calculate element internal forces, Fel=Kel del
Assembly global stiffness matrix Ks and internal force vector Fint

Calculate unbalanced forces Fext-Fint and global error
YES
Convergence
NO
YES
ite>IteMax
NO
Fig. 8 Flow chart of the solution algorithm
49
Secant reinforcing material stiffness is calculated similar to the relations (29):

f
Esi si
si
(33)
where subscript I means either x or y.

From the theoretical point of view, the stiffness matrix of the element is calculated with the
formula:
K el
D B dV
Vel
(34)
where B is the strain interpolation matrix. However, as it was stated at the beginning of the
chapter, there are closed forms to solve the integral (34) and no numerical integration is used.
From the element displacements, del, the new element stiffness matrix is evaluated as it was
already presented and then the internal forces are calculated with the formula:
Fel K el d el
(35)
The material laws are defined only for monotonic loading. Therefore, the solver used is not
capable to obtain the post-peak response of structures. However, there are two reasons why this
detail is not important:
i) Only maximum strength is important for the side walls of the socket;
ii) The presented formulation is based on stiffness matrix, which means that strain localization
will occur in post-peak response and the results significantly depend on the dimensions of
elements.
At each loading step, the vector of external forces is proposed based on the formula:
Fext FG FH
(36)
where:
FG are the gravitational forces;
FH represents the lateral force pattern proposed at the beginning of the analysis;
represents the scalar multiplier of the lateral force pattern that is to be increased at each
loading step.
To proceed with the analysis, a maximum and, respectively, a minimum scalar multiplier
increment should be declared. For one of the first loading steps, k, the scalar multiplier is:
k k max
(37)
However, the convergence is lost in one loading step, i. Yet, the structure may be capable to
carry larger loads than the acting ones. Therefore, the scalar multiplier for the next step is
decreased according to the formulas:
i _ new i _ old new

new 0.5 old 0.5 i 1 i 2
(38)
where:
i _ new , i _ old represent the new and respectively, the old scalar multiplier of the lateral
force pattern for pushover point i;
50
new , old represent the new and respectively, the old scalar multiplier increment of the
lateral force pattern;
i 1, i 2 represent the the scalar load multipliers for steps i-1 and i-2.
The procedure described is continued until the increment of the scalar multiplier is less than the
declared minimum scalar multiplier increment:
min
(39)
Convergence is considered to be lost when a prescribed number of iterations, IteMax is achieved

or when convergence is not uniform, which means that error is not reduced with an increasing
number of iterations. The error is calculated with the formula:
Err
Fext Fint
Fext
(40)
where |*| means the Euclidian norm of the vector.

The flow chart of the solution algorithm is presented in fig. 8.
The presented method has been implemented in a handmade program developed in MATLAB.
The program performs static analysis of 2D structures made from orthogonal membrane
elements. Input data is read from files. The program has a graphic user interface which permits
the visualization of the input data and the results. One of the results available, the cracking
pattern, is presented in fig. 9 for the side wall of a typical socket.
Fig. 9 Cracking pattern for the side wall of a typical socket
4. Analysis Results
The analysis followed the steps in the flow chart in chapter 2. The following parameters were
considered:
- the height of the columns: 7000mm;
- the normalized axial force of the columns: 0.1;
- the longitudinal reinforcement ratio of the columns: 0.01;
51
the sides of the square column sections, which vary from 350mm to 800mm, with 50mm
increment;
the Hs/Ls ratio, which varies from 0.8 to 1.6 with 0.05 increment;
the reinforcement which is designed according to the NP 112/2012 [3], except for the
horizontal tie which has been deliberately overdesigned;
the steel grade, which is BSt 500, while the concrete is shifted from C25/30 to C30/37
and C35/45;
the tension stiffening parameter in equation (17), which is considered 200;
the maximum dimension of the membrane elements, which is 200mm.
The aforementioned parameters produced three sets of 170 analysis results, each set
corresponding to a concrete grade. The strut stress strength is calibrated with the following
formula:
R 1 fcd
(41)
Additionally, the mean shear stress strength of the lateral wall of the socket was calculated with
the formula:
R 2 f ctd
(42)
where 1, 2 are parameters that depend on the concrete grade and the accepted probability of
failure.
Fig. 10, 11, 12 and 13 present the histogram for strut stress and mean shear stress strength for
C25/30 concrete. For the calibration of the resistance parameters, a log normal distribution was
considered, as inferred from the presented histograms.
Fig. 14 presents the coefficient of variation of the analyzed data. There are two main remarks:
i) the coefficient of variation is generally small (less than 0.20), which presumes high confidence
levels for the calibration parameters;
ii) the coefficient of variation for the strut stress strength is less than the coefficient of variation
for the mean shear stress strength, which confirms the superiority of the strut and tie model
proposed by NP 112/2012 [3].
Fig. 15 and 16 present the calibration coefficients for different accepted probability of failure for
the three concrete grades. It could be noted that there are small differences in the calibration
coefficients for different concrete grades. As a result, for design purposes it could be considered
that the calibration coefficients do not depend on concrete strength, but only on the accepted
probability of failure.
Table 1
Proposed design parameters
P[E<ER]
0.001
0.01
0.02
0.05
0.5
1.1
1.15
1.2
1.45
1.9
0.21
2.15
2.15
2.75
The proposed values of the calibration coefficients are presented in table 1. However, the
recommendation of the author is to use strut and tie model in the detriment of mean tangential
effort model for design purposes. It is notable that the calibrating parameters for strut stress
strength are larger than the ones proposed by code provision (0.75), even for small values of the
accepted probability of failure.
Taking into account that average concrete strengths were used in the study, the recommended
value of the accepted probability of failure is 0.1% for the design purposes.
52
39.4
C25/30
f [%]
30.0
13.5
8.8
5.3
11.21
12.48
13.76
15.03 16.31
s [MPa]
1.2
0.6
1.2
17.58
18.86
20.13
Fig. 10 Histogram of the strut stress strength, concrete C25/30
47.6
C25/30
f [%]
32.9
12.9
2.9
0.6
2.51
2.88
3.26
3.63
4.00
v [MPa]
1.2
0.6
1.2
4.38
4.75
5.13
Fig. 11 Histogram of mean shear stress strength, concrete C25/30

42.4
C30/37
f [%]
32.4
12.9
4.7
12.57
2.9
14.15
15.72
17.30 18.88
s [MPa]
1.8
1.8
1.2
20.46
22.03
23.61
Fig. 12 Histogram of the strut stress strength, concrete C30/37
53
44.7
39.4
f [%]
C30/3
7
5.9
3.5
2.91
3.36
3.81
1.8
1.8
1.8
1.2
4.26 4.71
v [MPa]
5.16
5.61
6.06
Fig. 13 Histogram of mean shear stress strength, concrete C30/37

0.17
0.11
0.15
0.13
0.12
0.12
C25/30
C30/37
C35/45
Fig. 14 Coefficients of variation of the analyzed data
1.50
1.47 1.45
C25/30
1.26
C30/37
1.22
1.17
1.20
1.17
1.11
1.17
C35/45
1.13
1.07
1.07
1.04
a1
0.97
0.5
0.05
0.02
0.01
0.001
P[s<sR]
Fig. 15 The calibration parameter of the strut stress strength for different accepted probability of failure
54
C25/30
2.69
2.78
C30/37
2.86
C35/45
2.25 2.29 2.28
2.16 2.18 2.15
2.09 2.11 2.07
a2
1.93 1.92 1.86
0.5
0.05
0.02
0.01
0.001
P[v<vR]
Fig. 16 The calibration parameter of the mean shear stress strength for different accepted probability of failure
5. Design Example
To illustrate the practical application of the suggested design procedure, the strut and tie model is
applied for the design example presented in Postelnicu [12]. The design data are:
- socket height: H s 1000mm ;
-
socket length: Ls 1100mm ;
socket thickness: tw 200mm ;

design shear force on the socket: VEd 1110kN ;
steel BSt 500, concrete C30/37.
The design steps are:

i) Calculate strut angle using relation (7):
tan
Hs
1000
1.11 48
Ls tw 1100 200
ii) Calculate horizontal tie reinforcement (fig. 17):

Ash, rq
0.5VEd 0.5 1110 103
1276mm 2
435
f yd
Horizontal tie reinforcement is distributed in the upper half of the socket wall. Eight horizontal
rebar 14 are used, four on each side of the wall, resulting in an effective area of 1232mm2
which is considered sufficient.
iii) Calculate vertical tie reinforcement (fig 17):
Asv, rq Ash, rq tan 1276 1.11 1416mm2
55
Vertical tie reinforcement is distributed in the corners of the socket. Four vertical rebars 22 are
used, resulting in an effective area of 1521mm2.
Fig. 17 The definition of the horizontal and vertical tie reinforcement
iv) Calculate the axial effort in the strut using equation (3):
Cw.Ed
0.5VEd 0.5 1110
830kN
cos cos 48
v) Calculate the strut width using equation (8):

bw tw sin 0.4 H s cos 200 sin 48 0.4 1000 cos 48
148.63 267.65 416mm
vi) Calculate the strut stress using equation (4):
w.Ed
Cw.Ed 830 103
10MPa
bw tw 416 200
vii) Calculate the strut stress strength for 0.1% failure probability using equations (6), (41) and
table 1:
w.R 1 fcd 1 0.6 1
30
20 10.6 MPa 10 MPa
250
It can be noted that the proposed wall thickness satisfies the strength requirements. Further
details of the socket design are presented in Postelnicu [12].
6. Conclusions
The paper is a further investigation of the strut and tie model proposed in the new code, NP
112/2012 [3] for the design of the side walls of the socket foundations. There are three main
improvements suggested for the code provisions:
i.
56
the definition of the strut inclination;

ii.
the change of the concrete strut width;
iii.
the change of the concrete strut resistance.
The proposed interventions provide a rational design approach which is calibrated based on a
simple, yet powerful, tool for the analysis of squat elements, MCFT. Strut stress strength is
calibrated for different accepted probability of failure, which provides design flexibility,
depending on design purposes. The calibration has an increased confidence level confirmed by
the small coefficients of variation of the analyzed data.
A future research direction is to study the strut resistance for different load distribution patterns.
Different load distribution patterns may appear, for example, if the frictional mechanism is
introduced in the design approach.
References
[1]
[1]
[2]
[3]
[4]
Leonhardt, F., Mnning, E. , 1973 Vorlesungen uber massivbau, Springer-Verlag.

NP 112, 2004 Normativ pentru proiectarea structurilor de fundare direct,
NP 112, 2012 Normativ privind proiectarea fundaiilor de suprafa, Redactarea II
EN 1992-1-1, 2004 Design of concrete structures. General rules and rules for buildings, CEN
Vecchio, F. J., Collins, M. P. March-April 1986 The Modified Compression Field Theory for Reinforced
Concrete Elements Subjected to Shear, in ACI Journal, Vol. 83, No. 2, pp. 219-231
[5] Collins, M. P., Mitchell, D., 1987 Prestressed Concrete Basics, Canadian Prestressed Concrete Institute
[6] Bentz, E. C., 2000 Sectional Analysis of Reinforced Concrete Members, PhD Thesis, University of Toronto
[7] Vecchio, F. J., March 1990 Reinforced Concrete Membrane Element Formulation, in ACI Journal, Vol. 116,
No. 3, pp. 730-750
[8] Criesfield, M. A., 1981 A Fast Incremental/Iterative Solution Procedure that Handles Snap-Through,
Computers and Structures, Vol. 13, pp. 55-62
[9] Bathe, K. J., Dvorkin E. N., 1983 On the Automatic Solution of Nonlinear Finite Element Equations,
Computers and Structures, Vol. 17, No. 5-6, pp. 871-879
[10] Powell, G., Simons, J., 1981 Improved Iteration Strategy for Nonlinear Structures, in International Journal
for Numerical Methods in Engineering, Vol. 17, pp. 1455-1467
[11] Postelnicu, T., 2013 Proiectarea structurilor de beton armat n zone seismice, Editura MarLink, Bucureti,
57
CONCEPTION OF AN ADVANCED THERMAL MANIKIN FOR

THERMAL COMFORT ASSESSMENT IN BUILDINGS AND VEHICLES
Angel DOGEANU-PhD student, CAMBI research center, Technical University of Civil Engineering Bucharest,
Faculty of Building Services
Ciprian CALIANU-Undergraduate Student, Technical University of Civil Engineering Bucharest, Faculty of
Building Services
George CHITARU-Undergraduate Student, Technical University of Civil Engineering Bucharest, Faculty of
Building Services
Matei GEORGESCU-Undergraduate Student, Technical University of Civil Engineering Bucharest, Faculty of
Building Services
Andrei TUDORACHE-Undergraduate Student, Technical University of Civil Engineering Bucharest, Faculty of
Building Services, e-mail: angel.dogeanu@yahoo.com
Abstract: This paper is presenting one of the five prototypes of thermal manikins conceived at the
Building Services Faculty (Thermal-Hydraulic Systems Laboratory) at the Technical University of
Civil Engineering. This particular prototype has an advanced anatomic shape, with 36 independent
active zones, temperature sensors and its own in-house software for data acquisition and control of the
zone surface temperature. The paper is presenting all stages needed for the development of an
advanced thermal manikin for research purposes. All composing parts of the manikin and the
validation strategy were presented. A carefully check-up of all components was necessary and some
of the tests we done were time consuming.
The model is fully functional and the temperatures of each body zone can be easily modified in
accordance to our needs.
Keywords: thermal manikin, thermal comfort, CFD validation
1. Introduction
Thermal manikins were used for more than 70 years. At the beginning they were used for testing
clothing for soldiers by the US Army [1]. The shape and heating system were very simple at that
stage. Since then the shape and complexity of the thermal manikin has enhanced and started to
approach the complexity of the human body. The number of independently controlled zones
increased from a single zone corresponding to the entire surface up to 120 individually
controlled zones [2]. The material used for developing the thermal manikin have diversified,
from copper to plastic and carbon fibber to skin like silicon. In the earlier stages the purpose of
the thermal manikin was limited to clothing testing. In time the range of applications has become
wider. The use of thermal manikins in the field of thermal comfort research has become more
and more active and the technology used to create them has now different approaches. Most of
them try to simulate the human body and the heat emission in the environment, while others are
more or less complex measurement devices for assessing thermal environment quality by
simulating the human body thermal regulation mechanisms and measuring its heat loss towards
its environment [1]. The most advanced of them can also simulate body sweating and heat
exchange through evaporation [1-4].
This paper is presenting one of the five prototypes of thermal manikins conceived at the Building
Services Faculty (Thermal-Hydraulic Systems Laboratory) at the Technical University of Civil
Engineering. The first prototype had an anatomic shape of the body and 6 zones made of heating
ribbon and was developed during the PhD thesis of Dr. Cristiana Croitoru [5]. A second and a
third thermal manikin have a square simplified shape with respectively 3 [6] and 9 [7] zones
made of heating polyamide film. A fourth prototype under construction regards an advanced
thermal breathing patient for healthcare facility ambient comfort assessment [8]. The prototype
presented in this paper has an advanced anatomic shape, with 36 independent active zones,
temperature sensors and its own in-house software for data acquisition and control of the zone
surface temperature.
58
2. Architecture of the thermal manikin

The architecture of the thermal manikin as a heating source and a measurement device is
presented in Figs. 1 and 2. Its working principle is to control the surface temperature of each
individual zone and to record the electrical power consumption as an indication of the thermal
state of each zone. Each heating circuit (i.e. zone or segment of the manikin) is heated as a
function of the indication of a process controller which in its turn depends on the temperature
recording of the sensors placed on each zone. In our case there are 36 individually controlled
heating zones, each of them provided with two temperature sensors. A dedicated software
interface is allowing the user to specify set-point values for each surface temperature of the
zones, to monitor the evolution of each temperature of the 72 available sensors and to record the
electric power consumption of the segments.
The global working principle of the manikin is presented in Fig 1, the software interface is
displayed in Fig. 2 and the architecture of the control interface of the manikin is presented in Fig. 3.
Fig 1 - Working principle of the thermal manikin
Fig 2 - Data acquisition and control software interface
Fig 3 - Architecture of the control part of the manikin
59
3. Building the heating part of the thermal manikin

The thermal manikin was designed for both seated and standing postures. It is divided in 36
zones as displayed in Fig. 4 and 5. The size of the manikin is a standard human size with a total
surface of 1.8m (Fig. 5). The base structure of the manikin is made of polyurethane foam placed
on a metallic skeleton.
The heating solution chosen for this prototype consist in using five elementary patches (Fig. 6a)
that are combined to cover each individually controlled zone. The patches are made of a thin
layer of silicone (1.5mm) that includes a heating circuit made of nickel chrome heating wire and
etching foil. The heating silicone patches are presented in Fig. 6.
Several solutions of films and heating materials were tested but the heating silicone patches were
found to provide the best uniformity in terms of temperature distribution. In Fig. 4 the surface
temperature distributions of a flexible polyamide heating film, of a silicone patch and of the
human skin of the hand are compared.
The heating film patches were placed on the polyurethane base using double side adhesive tape.
After covering a body zone the electrical connections and circuits were created. For safety
reasons, every electrical connection was tested for leakage. The electrical wire was embedded in
the polystyrene basis (Fig 8a). In order to ensure that the thermal load of the film mounted over
the wires is not influencing the cable stability we selected special electrical wire that works at
temperatures above 7C.
During a preliminary test, without any control of the circuits, the temperature of each zone
stabilized at 45C when the room temperature was stable at 24 C, a rather encouraging result
offering a wide range to control the temperature of each zone and the possibility to simulate
different cases of body heat release.
Fig.4 - Independent controlled body parts (bordered by yellow tape)
60
Fig.5 - Distribution of the anatomic zones individually controlled
a)
b)
c)
Fig. 5 - a) and b) Photographs of the heating silicon parts that are composing the zones,
c) Sketch of the composition of the heated film

61
a)
b)
Fig. 6 - a) Non uniform distribution of the temperature field on a flexible polyamide heating film, b) Comparison
between the surface temperature distribution of a silicone patch and the human skin of the hand.
Even if this type of heating element appeared to be the best possible solution for the current
project there are still some regions were the heating circuit is not present creating zones were the
temperature values are lower (i.e. the borders of each patch see Fig. 6 b for instance). As a nonuniform temperature of the manikin surface has to be avoided [9] we decide to cover the entire
manikin surface with adhesive aluminium foil to ensure enhanced conduction heat transfer.
Finally, in order to facilitate further investigations with a thermal (IR) camera the entire manikin
surface was covered with a transparent adhesive film. The manikin will be covered by real
clothes according to the studied situation or with a light black textile.
In order to validate the heating part of the manikin prototype we performed several tests,
concerning the distribution of the surface temperature of each manikin, the sensors employed for
the control part, the electrical characteristics of the used components.
a) Tests of the surface temperature distribution
For this test we used an infrared camera FLIR E60. During the tests the air temperature was
stable at 24C. Each zone was electrically powered and checked for its integrity and its thermal
response over time (Figs. 7-9). The surface temperature control of each zone was performed
using a sophisticated programmable power source HAMEG HMP-2020 (Fig.7a). The
temperature for each body zone was set using the measurements made with the IR camera and
the K-type thermocouples controlled through a National Instruments acquisition board and
Labview Software (Fig.7b).
a)
b)
c)
Fig. 7 - a) Programmable power source used during the tests of the heated zones of the manikin, b) Data acquisition
board, c) Labview software
62
This high performance measurement facility allowed us to follow the real time response of our
heating elements to different thermal solicitations or to calibrate the IR camera, to record the
temporal evolution of the temperature gradient over a surface, etc.
a)
b)
Fig. 8 - Preliminary tests for checking the uniformity of the surface temperature of a zone composed of several
silicone heating patches: a) set-up of termocuples, b) infrared thermography images
An example of the performed tests is presented in Fig. 8 where we checked the surface
temperature uniformity of a single zone composed of two silicone heating patches. In this case
we wanted to evaluate the response of the heating elements in the junction zone (i.e. the border
zone where two additional patches are joined together). These preliminary tests helped us to
choose the positions of the digital temperature sensors for the control part of the manikin.
b) Tests of the temperature sensors
For the control part of the surface temperature of the manikin we choose digital sensors given the
adopted strategy presented previously. The sensors are the model TSic 706F manufactured by
IST. Each sensor was carefully tested and calibrated in our laboratory. In Fig.12 a and b
examples of response at thermal solicitations over time for eight of these sensors with and
without calibration are presented. During the calibration process we used a thermostatic water
bath Lauda Eco with immersion thermostat Lauda Eco Silver (Fig. 12 c).
a)
63
b)
Fig. 9 - a) Different phases in the made-up of the thermal manikin, b) testing each zone of the manikin
Fig. 10 - Comparisons between the temperature distributions of different body parts of the manikin and those of a
human body
Fig 11 - TSic 706F sensors used for the measurement and control part of the thermal manikin
64
45
45
withoutcalibration
40
withcalibration
S1
S1
40
S2
S2
[C ]
S3
35
S4
[C ]
S5
30
S3
35
S4
30
S5
S6
S6
S7
25
25
S7
S8
S8
20
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
reference
probe
[min]
20
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
[min]
reference
probe
Fig 12 - Response at thermal solicitations over time of TSic TO92 sensors:

a) without calibration, b) with calibration;
c) thermostatic water bath Lauda Eco with immersion thermostat Lauda Eco Silver
c)
Electrical power consumption tests
The electrical power is the only power source that at is converted in heat with efficiency near to 1.
During preliminary tests of electrical power consumption we aimed to asses an evaluation of the
electrical consumption of the thermal manikin [1]. In order to do this we used a high performance
wattmeter, we measured the power consumed by the wattmeter without any consumer and found
that it consumes 2W. Afterwards for each heating zone of the manikin we evaluated the power
consumption in permanent regime for certain values of the set-point temperature and for the same
fixed value of the ambient air temperature.
Other tests evaluated the electrical power consumption under limit conditions such as a local cooling
of the manikin body parts or the extreme ambient air conditions at which the manikin was designed.
As a design condition the manikin surface is 1.8 m which is the equivalent surface of standardized
human body [10]. After the electrical power measurements we find the specific heat flux (i.e.
W/m) generated by the thermal manikin to have coherent values equal to the dry heat generated by
a human body in several types of activities.
These tests allowed us to choose the best suited electrical power source. We chose a Mean Well
RSP power source of 2.4 kW.
65
3. Building the control part of the thermal manikin

The control part of the thermal manikin is following the protocol presented in Figure 3. It is
composed of two parts: the data acquisition and monitoring part and the heating control part.
The data acquisition and monitoring part has to comply with several demands, such as: having
the less possible number of connecting cables, having the less number of perturbed elements
between the manikin itself and the evaluated thermal ambient, the fastest acquisition rate.
The data acquisition system centralizes data from 72 sensors and transmits them to the control
part. This transmission is done via a Bluetooth communication module which has the advantage
of a rapid frequency of data communication (up to 2.4 GHz). Each temperature sensor
communicates through a serial interface, after a time response of 70-100ms from its activation.
In order to avoid a very slow response of the entire system, we chose to create a distributed
architecture of the data acquisition system in order to parallelize processes (Fig 13). In this case
eight distinct data acquisition modules were created (Fig 13 and 15 a) allowing a theoretical
sampling frequency of 1Hz.
The topology of the data acquisition system is of the token-ring type. The modules are
receiving messages from only one neighbour and can send messages to only one neighbour in only
one direction of the ring (Fig 13). A master module (main communication module) initiates data
communications ensuring that links between modules are operational and working properly,
allocating software identities to each slave module. Once all data received, they are communicated
through the Bluetooth Module to the main control unity. The latter commands the start and the stop
of the acquisition system and can interrogate upon some working parameters if necessary.
Fig 13 - Architecture of the data acquisition part
66
Fig 14 - Architecture of a slave module
Each of the slave modules waits the initialization command from the master module and then
starts data acquisition from the eight dedicated sensors. Once the data are acquired, it will wait
for the neighbour slave module to send the acquired data and then to append its own values of
temperature or any error message if applicable. All the data sequence is then sent to the other
neighbour module that is waiting. The cycle is repeated for a new data acquisition run.
The main control system communicates through the Bluetooth module with the data acquisition
part and has access to all temperatures of all temperature sensors on the manikin surface. Once
the data are received, the main control system may call the regulation routine for maintaining the
surface temperature constant. At the same time it can interact with the user via the software
interface on a dedicated PC (Fig 3). The user indicates the set-point temperature for each zone
and may follow the real time evolutions of each zone surface temperature and power
consumption.
a)
b)
c)
Fig 15 - In house developed control boards for the thermal manikin: a) control of the sensors, b) control of the
power source, c) blue tooth communication control
67
5. Conclusions
This paper is presenting all stages needed for the development of an advanced thermal manikin
for research purposes. All composing parts of the manikin and the validation strategy were
presented. A carefully check-up of all components was necessary and some of the tests we done
were time consuming.
The model is fully functional and the temperatures of each body zone can be easily modified in
accordance to our needs.
Acknowledgements
This work was supported by the grants of the Romanian National Authority for Scientific
Research, CNCS UEFISCDI, project numbers: PN-II-ID-PCE-2011-3-0835 and PN-II-PTPCCA-2011-3.2-1212.
References
[1]. Holmr, I., Thermal manikin history and applications. European Journal of Applied Physiology, 2004. 92 p.
614-618.
[1] Rugh, J.P. and D. Bharathan. Predicting Human Thermal Comfort in Automobiles in Vehicle Thermal
Management Systems Conference and Exhibition,. 2005. Toronto, Canada
[2] Alahmer, A., et al., Vehicular thermal comfort models; a comprehensive review. Applied Thermal Engineering,
2011. 31(67): p. 995-1002.
[3] Sakoi, T., et al., Thermal comfort, skin temperature distribution, and sensible heat loss distribution in the sitting
posture in various asymmetric radiant fields. Building and Environment, 2007. 42(12): p. 3984-3999.
[4] Croitoru, C., Studii teoretice i experimentale referitoare la influena turbulenei aerului din ncperile
climatizate asupra confortului termic, in UTCB. 2011.
[5] Cruceanu, I.A., Studiul micrii aerului n sistemele de ventilare personalizate, Thesis, UTCB. 2013.
[6] Dogeanu, A., et al., Conception of a simplified seated thermal manikin for CFD validation purposes. Romanian
Journal of Civil Engineering, 2013(december 2013).
[7] EQUATOR - Grant of the Romanian National Authority for Scientific Research, CNCS UEFISCDI, project
number: PN-II-PT-PCCA-2011-3.2-1212.
[8] Melikov, A.K., Breathing thermal manikins for indoor environment assessment: important characteristics and
requirements. European Journal of Applied Physiology, 2004. 92(6): p. 710-713.
[9] Wang, J. and E. Hihara, Human body surface area: a theoretical approach. European Journal of Applied
Physiology, 2004. 91: p. 425-428.
68
ANALYSIS OF CREST CUTOFF WALL AT OSTROVUL MIC LEFT SIDE

EMBANKMENT DAM
Daniel Andrei GAFTOI Assistant, PhD, Technical University of Civil Engineering, Hydrotechnics Engineering
Department, e-mail:daniel_gaftoi@yahoo.com
Ctlin POPESCU - Assistant, PhD, Technical University of Civil Engineering, Hydrotechnics Engineering
Department, e-mail:catalin.popescu@utcb.ro
Drago FRILESCU PhD Student, Technical University of Civil Engineering, e-mail: dragos.fratilescu@yahoo.com
Abstract: Ostrovul Mic hydropower plant, the first one of the Rul Mare Aval scheme, creates a
reservoir controlled by a gated dam and bordered by 2 km of embankment lateral dams made of
river ballast and watertight on the upstream face by a reinforced concrete slab and, within the
foundation, by a cut-off wall at the upstream toe. The lateral dams, with a maximum height of 20
m, faced seepage problems since commissioning in 1986 and this is the cause for restricted water
level operating mode most of the time. Over the years a series of remedial works were realized, but
they failed to effectively control the seepage flow and currently the seepage phenomena evolved:
wet area has extended, the number of springs has raised and the level of emergence has also risen.
This situation led to the necessity of analyzing a series of remedial works and the current paper
presents the detailed analysis realized on a mathematical model for one of the three remedial
alternatives proposed - an impervious cutoff wall realized from the dam crest and embedded into
bedrock. The purpose for the analysis was to determine the performance of the proposed solution in
terms of seepage flow, seepage gradient, level of emergence on downstream face and horizontal
influence of the cutoff wall length.
Keywords: seepage analysis, mathematical modeling, cutoff wall, embankment dike, finite
element method
1. Introduction
All fill dams face seepage through the embankment or through the foundation as the impounded
water seeks paths of least resistance. These phenomena, left uncontrolled and accompanied by
hydrodynamic picking up of the fine material, threaten dam stability and may cause unforeseen
failure. The ground conditions and the geological features of the dam site greatly influence the
amount of seepage and its relevant effects [1] [2].
Such phenomenon is also noticed at the Ostrovul Mic hydropower development, where strong
seepage associated with fine material picking up through the perimeter dams of the lake, mostly
through the left side, is present. Due to this phenomenon, which may lead to dam instability, the
water level is restricted and this is causing important energy and financial loses.
The paper presents, at the beginning, a short description of these phenomena, followed by a 2D
vertical and horizontal plane analysis of one of the 3 proposed remedial solutions a crest cutoff
wall. As a result of the analysis, the efficiency of the proposed solution regarding the seepage
flow, the spring level, the critical gradient and the proposed execution length influence were
determined.
2. Site description
Ostrovul Mic hydropower development (figure 1), first of the Rul Mare Aval scheme, is
situated at about 20 km from Haeg town. The development is composed of a weir dam, a
hydroelectric power plant and the perimeter dams. The main purposes of the development are:
the production of electrical energy, the regulation of the Raul Mare River and irrigation. The
normal retention level (NRL) of the lake is 465,00 maSL.
69
Fig. 1 Ostrovul Mic HPP - Plan view and panoramic photo

With red the interest area (left side dike at km 1+230 1+370)
Legend: 1. Spillway dam, 2. Stilling basin I, 3. Stilling basin II, 4. River bed, 5. Hydro Power Plant, 6 Outlet
channel, 7. Tailrace, 8. Concrete face, 9. Vegetated slope
The embankment dikes are made of ballast with trapezoidal cross-section having a 4.00 m crest
width, 1:2 upstream slope and 1:2 downstream slope for dike height higher than 20.00 m and
1:18 otherwise. The sealing system consists of a reinforced concrete mask made of 4.00 m x 5.00
m plates of 15-20 cm thickness, with bituminous sealed joints and a 5-6 m depth sealing cutoff
wall made as a trapezoidal ditch filled with concrete gel.
In order to permit the embedding of the trapezoidal cutoff wall into the badrock, the ditch was
realized under the terrain level, from a 4 m excavation made at the dikes toe (figure 2).
Fig. 2 - Left side dyke cross section
Following the filling of the lake, due to the chosen solutions and the encountered difficulties, the
sealing system has the following shortcomings often: the cutoff wall is not embedded into the
bedrock, there is no joint between the cutoff wall and the cutoff support beam and there are
windows in the cutoff wall in the lower area of large blocks.
70
3. Atypical phenomena during lake operation Causes and proposed remedial solutions
The dikes of the Ostrovul Mic development have faced seepage problems since its
commissioning in 1986, problems which represent the main cause of the level restrictions for
most of the working time. Over time, different remedial works were done, but with limited
effects in time. At this moment, the most affected part of the dike is situated on the left dike at
1+300 km.
Even though during the last years, the lake water level has been restricted to 462.50 maSL, in the
mentioned area the number of the concentrated springs has raised, the wet zone and the spring
level have grown, the measured seepage flow have reached approximately 900 l/s.
The main cause for this situation has been considered to be the existent cutoff wall defects (such
as the lack of embedding of the cutoff wall into the bedrock and the presence of windows in the
areas inferior to large blocks), but as a result of the calibration calculus, presented in the
following parts of this paper, we may demonstrate that they represent only a secondary cause.
The main cause of the large seepage flows and the wet areas on the downstream slope is the
degradation of the concrete mask in the inferior area of the dam and the lack of junction between
the cut off wall and the beam.
In order to solve these problems, 3 remedial solutions of the dike have been proposed, the
present paper only summarizing the vertical and horizontal plane analysis for the first one:
- 1st solution: the construction of a cutoff wall from the crest of the dike along its vertical
axis, embedded in the bedrock at least 1 m
- 2nd solution: the construction of an additional cutoff wall, in front of the existing one,
from the crest to the bedrock and the rehabilitation of the concrete mask
- 3rd solution: the remaking of the superior sealing and the joint between the concrete mask
and the cutoff wall using a geomembrane placed on the existing mask
4. Crest cutoff wall analysis
4.1. Mathematical models used
The analysis of the crest cutoff wall effect was realized both in the vertical and horizontal plane
using 2 steady state mathematical models based on the finite elements method. The analyses
were done using Geoslope 2004 software, Seep/w module.
For the 2D vertical plane analysis, the left dike embankment cross section was represented as a
mesh consisting of 12580 triangular and quadrilateral finite elements. The permeability values
for the analyzed materials were determined during the model calibration phase, starting from the
values from the geotechnical studies.
The 2D horizontal plane analysis included the dike, the channel and the nearby areas along the
300 m long sealing proposed length and also 150 m long upstream-downstream area. The
mathematical model consists of a mesh of 4579 triangular and quadrilateral finite elements
connected by 4549 nodes. The transmissivity of the analyzed areas was determined based on the
permeability values used in the 2D vertical plane calculus.
As boundary conditions the following elements were considered:
- Upstream: the lake water level is equal to the restricted level (462.50 maSL) for the calibration
phase and to the normal retention level (NRL: 465.00 maSL) for the analysis phase
- Downstream: the downstream channel water level is equal to 1 m above the level of the deck
of the downstream channel (449.00 maSL) and the potential spring area on the downstream
slope.
71
In the first phase - calibration - the model included the body of the dike with the existing sealing
systems and foundation ground, and in the second phase - analysis - the model was modified
according to each analyzed alternative.
4.2. Mathematical model calibration
In order to determine the permeability values of the analysis materials, the emergence elevation
of the downstream seepage curve was chosen as an element of calibration for the model. To
determine its value in situ, infrared photography was used, which, according to Stematiu et al.
[3] and Moel et al. [2], offers a fast, simple, flexible and nondestructive inspection of the zones
where the longitudinal dams are not equipped with piezometers or other behavior monitoring
systems. The realized investigations and the way the emergence elevation of the seepage curve is
determined are summarized in figure 3.
Fig. 3 Seepage line determination
Comparing the digital photography with the infrared photography, with the silhouette of the
witness as the connecting element for the two pictures, we can affirm that the blue/red
demarcation corresponds to the depression curve emergence zone approximately 456.00 maSL.
Starting from the values determined in laboratory and using an iterative process, the permeability
and anisotropy values were modified in order to obtain an emergence rate of the depression
curve similar to those in situ (table 1). The calibration was performed for two distinctive cases,
one considering the cutoff wall embedded into the bedrock, the other one considering a 0.5 m
gap between the cutoff wall and the bedrock.
Table 1
Permeability values for the calibrated model
Material
Dike body (filling)
Alluvium foundation
Bedrock
Deteriorated concrete mask
Existing cutoff wall
New cutoff wall
72
Kx
m/s
2 x 10-3
4 x 10-3
1 x 10-6
2 x 10-4
1 x 10-6
1 x 10-6
Ky
m/s
2 x 10-2
4 x 10-3
1 x 10-6
2 x 10-4
1 x 10-6
1 x 10-6
The emergence level on the downstream face shown by the flow net in figure 4 corresponds to
the measured one. Similar values were obtained in both cases embedded and floating cutoff
wall thus revealing that the main cause of the recorded seepage phenomena is not the
deficiency of the performed foundation sealing. As mentioned before, the simulations have
pointed out that the most significant contribution to seepage and to the high level of seepage line
are done by the deteriorated slabs on the lower part of the concrete face and especially by the
damaged joint between the concrete face and the cutoff wall. The excavations in front of the
upstream toe, required to implement the cut-off wall in an open trench, have exposed the most
pervious layer above the bedrock to direct infiltration.
Fig. 4 Flow net for the calibrated model:
a concrete face deteriorated, cutoff wall not embedded; b concrete face deteriorated, cutoff wall embedded
4.3. 2D vertical plane analysis

In the second phase, the crest cutoff wall efficiency was analyzed and the calibrated model was
geometrically modified so it would simulate the effect of the sealing works.
The effect of these works was analyzed both in the situation of a perfect execution (figure 5) and
in the situation of an execution with defects. The analyzed defects were: lack of embedding in
the bedrock (figure 6) and the presence of windows in the body of the cutoff wall (figure 7).
480
475
470
Cota coronament : 466.80 mdM
Cota lac: 465.00 mdM
465
Cote
460
455
450
445
440
458
462
460
452
456
454
430
0
45
46
4
435
425
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
110
120
130
140
150
Distante orizontale
Fig. 5 Flow net perfect execution

480
475
470

465
Cote
460
455
450
445
440
435
430
425
10
20
30
40
50
60
70
80
90
100
Distante orizontale
Fig. 6 Flow net lack of embedding 0.50 m
73
480
475
470
465
Cote
460
455
450
445
458
462
460
452
456
430
450
435
454
464
440
425
0
10
20
30
40
50
60
70
80
90
100
110
120
130
140
150
Distante orizontale
Fig. 7 Flow net window 0,50 m
For an easier comparison, table 2 summarizes the seepage flow, the elevation of the emergence
of the seepage curve above the berm, the flow rate, the wet area on the downstream slope and the
relation between exit gradients and the critical one determined based on the granulometric curves
of the filling material. Also, the first two rows of the table show the current state results, without
remedial works, presuming the retention level reaches the normal water level.
Table 2
Simulation results of the analyzed hypotheses
Hypothesis
Current state (deteriorated concrete mask,
cutoff wall not embedded)
Current state (deteriorated concrete mask,
embedded cutoff wall)
Crest cutoff wall perfect execution
q_total
l/s,ml
Spring level
measured from the
berm, height
Wet area
m2/ml
i < icritical
0,992
9,58 / 460,08
19,74
No
0,986
9,58 / 460,08
19,74
No
Yes
Yes
Yes
0,99
Yes
0,037
Crest cutoff wall lack of embedding (0,50 m)

Crest cutoff wall with the presence of a window of
0,25 m
Crest cutoff wall with the presence of a window of
0,50 m
0,419
0,392
0,494
0 / into the side

channel
0 / into the side
channel
0 / into the side
channel
0,48 / 450,98
The analysis of the results shows that the crest cutoff wall offers a complete seepage control. In
the event of a perfect execution, without windows and with the embedding realized in the
bedrock, the seepage curve sets off in the side channel, the wet areas in the downstream
parameter no longer exist, and the seepage flow in the critical area is insignificant.
In case of an execution with defects due to foundation granulometry - which presents large
boulders - the seepage curve sets off in the side channel, the hydraulic gradients are inferior to
the critical ones and the seepage flow is at most 0.5 l/s m. The springs located above the berm
elevation for the last hypothesis are insignificant.
4.4. 2D horizontal plane analysis
In order to analyze the effect of the cutoff wall in the horizontal plane and to determine its influential
area, a 2D horizontal plane model was made. Using this model, 4 different cases were studied:
- the dike operation under ideal conditions (hypothesis used for the results comparison)
- the effect of the cutoff wall, assuming it has no defects and a length of 300 m
(representing the first proposed remedial length)
- the effect of the cutoff wall, assuming it has defects (a 1 m x 4 m window)
74
the execution of a 150 m long cutoff wall
and the results (general view and details) are presented in the figures 8 14, as flow nets and
flow vectors.
The dike operation under ideal conditions basic hypothesis (used for the results comparison)
170
160
150
140
130
Distante (Y)
120
110
100
90
80
70
60
50
40
30
20
10
0
25
50
75
100
125
150
175
200
225
250
275
300
325
350
375
400
425
450
475
500
525
550
575
600
625
Distante (X)
Fig. 8 Flow net and flow vector basic hypothesis
This hypothesis was studied in order to be able to analyze the effect of the crest cutoff wall. As
expected, the equipotential lines of the flow net are parallel and the flow vectors are
perpendicular on the downstream channel.
The effect of the cutoff wall (perfect execution)
170
160
150
140
130
Distante (Y)
120
110
100
90
80
70
60
50
40
30
20
10
0
25
50
75
100
125
150
175
200
225
250
275
300
325
350
375
400
425
450
475
500
525
550
575
600
625
Distante (X)
Fig. 9 Flow net and flow vector crest cutoff wall (perfect execution)
This hypothesis studied the effect of a perfect execution of the cutoff wall. It is visible that the
flow net gathers at the ends of the cutoff wall, and the maximum flow vectors are also
concentrated in the same area. Figure 10 presents a detailed view of one of the cutoff wall end
areas. The analysis of the results showed that the effect of the screen is felt on approximately 50
m upstream-downstream, with a uniform flow through the rest of the dike.
Fig. 10 Flow net and flow vector crest cutoff wall (perfect execution). Detailed view of the cutoff wall end area
75
The effect of the cutoff wall, assuming it presents a defect situated at a from one of its sides
170
160
150
140
130
Distante (Y)
120
110
100
90
80
70
60
50
40
30
20
10
0
25
50
75
100
125
150
175
200
225
250
275
300
325
350
375
400
425
450
475
500
525
550
575
600
625
Distante (X)
Fig. 11 Flow net and flow vectors crest cutoff wall with defect
This hypothesis studied the effect of a crest cutoff wall that presented a defect during execution.
The presumed defect consists of a 0.5 m deep and 4 m long window inside the body of the wall,
situated at a from one of its sides.
Both figure 11 and figure 12 show that the effect of the cutoff wall in the upstream area (where
the defect is present) is insignificantly reduced, as the model presents no flow in the middle area.
Compared to the perfect cutoff wall hypothesis, the flow net does not change, as the lines still
gather at the end of the wall. The only line that visibly changed its trajectory is the 452 m line,
this change marking the effect of the presumed window.
45
2
45
8
454
456
46
0
462
454
464
Fig. 12 Flow net and flow vectors crest cutoff wall with defect
Detailed view of the cutoff wall end
The effect of a cutoff wall with reduced length ( length of the analyzed screen)
170
160
150
140
130
Distante (Y)
120
110
100
90
80
70
60
50
40
30
20
10
0
25
50
75
100
125
150
175
200
225
250
275
300
325
350
375
400
425
450
475
500
525
550
575
600
625
Distante (X)
Fig. 13 Flow net and flow vectors 150 m long crest cutoff wall
76
This hypothesis studied the execution of a 150 m long crest cutoff wall situated in the area with
the most important sealing defects, in order to analyze the effect of a reduced length wall (as
shown in figure 13 and 14).
It is visible that the length cut of the wall has an important effect on the protected area. Unlike
the previous hypotheses in which the middle area did not present any flow vectors, in this case
the water flows also through the middle of the dike. Regarding the influence of the wall on the
water flow through the rest of the dike (the area without a cutoff wall), no changes are visible
and the effect is also felt for approximately 50 m upstream- downstream.
45
4
45
452
458
462
464
45
6
458
45
4
460
460
462
464
Fig. 14 Flow net and flow vectors Detailed view of the 150 m long crest cutoff wall
5. Conclusions
The seepage phenomenon through the dikes of the Ostrovul Mic Lake is a continuous and
intensity increasing process, as the water is infiltrating through preferential paths, followed by
the fine material picking up. This is the cause of the restricted water level operating mode since
the commissioning of the development.
The seepage analysis was based on a finite elements model using Geostudio software. The
calibration of the model was realized using the emergence rate of the downstream parameter
depression curve. To determine its value in situ, infrared image technique was used.
The proposed solution for the sealing of the left side dike crest cutoff wall was analyzed in
both the situation when the execution was perfect and the situation in which some defects may
appear in its body. As a result of the mathematical modeling, this solution offers a complete
seepage control, even if the execution of the cutoff wall is not perfect.
6. Acknowledgements
Thanks are expressed to the Institute of Hydroelectric Studies and Design for the opportunity to
work on this project and to Professor Dan Stematiu for the valuable comments and exchange of
ideas that contributed to the accomplishment of this analysis.
References
[1] Uromeihy, A., Barzegari, G. Evaluation and treatment of seepage problems at Chapar-Abad Dam, Iran,
Science Direct, in Engineering Geology, No 91, 2007, pp. 219-228
[2] DeMoel M., Kodikara J., Adikari G. Seepage detection within embankment dams using infrared thermal
imaging, ANCOLD Annual Conference, 2010
[3] Stematiu D., Srghiut R., Popescu C., Gaftoi D Investigation techniques and remedial works to control
seepage through an earthfill dam, International Symposium on DAMS FOR A CHANGING WORLD -Need for
Knowledge Transfer across the Generations & the World, pp. 5-49 5-52, 2012
[4] Gaftoi D. - Seepage control through dams, PhD Thesis, Technical University of Civil Engineering, Bucharest,
2012
[5] Krahn, J. Seepage modeling with SEEP/W, User Guide, 2004
77
ENERGY SAVING ANALYSIS INSIDE A DOUBLE SKIN FACADE

Sebastian HUDITEANU - Ph.D. student, Technical University Gheorghe Asachi of Iasi, Faculty of Civil
Engineering and Building Services, email: seby_hudisteanu@yahoo.com
Claudia-Florentina POENARI - Master student, Technical University Gheorghe Asachi of Iasi, Faculty of Civil
Engineering and Building Services, email: poenari_claudia@yahoo.com
Bogdan-Iulian BALINT - Master student, Technical University Gheorghe Asachi of Iasi, Faculty of Civil
Engineering and Building Services, email: blnt_bogdan@yahoo.com
Monica CHERECHE - S.R. 3rd degree, PhD, Physicist, URBAN-INCERC, Hygrothermal - Climatic Research
and Testing Laboratory HT, Iasi, email: putina_monica@yahoo.com
Nelu-Cristian CHERECHE* - Lecturer, PhD, Technical University Gheorghe Asachi of Iasi, Faculty of Civil
Engineering and Building Services, email: chereches@tuiasi.ro
Abstract: This paper presents the energy savings of a room by using a double-skin faade system,
a concept that became an important architectural element for office buildings in the last two
decades. The results present a comparative analysis between the energy consumption of the room
without double skin faade and of the proposed case. The study presents the numerical simulation
performed in ANSYS-Fluent software. The numerical case was validated by experimental results
from [6]. The energetic and economic evaluation is realized by using the specialized software
RETScreen where the energy consumption and the rate of return on investment are highlighted.
Keywords: double-skin faade, situ, numerical modeling, energy consumption
1. Introduction
The double-skin faades are part of transparent faades [1]. Their goal is to keep as many of the
advantages of natural lighting and superior aesthetics and simultaneously removing some of
disadvantages such as low thermal resistance, noise exposure or overheating during the warm season.
a) natural ventilation
b) mechanical ventilation
Fig. 1 - Components of double-skin faade system [1]
The structure and the components of these systems are shown in Fig. 1. Ventilated double-skin
faades are in fact a simple improvement of the faade at a distance from the location of a glass
layer on the outside of the building. In this way it creates an air channel between the existing
glazing and the location. This concept can be successfully applied both to existing and new
buildings.
78
The elements in Figure 1 have the following meanings: 1 - exterior glazing, 2 - interior glazing,
3 - air channel, 4 - devices inside the channel; 5 - input and output sections, 6 - discharge pipe.
2. Numerical modeling
2.1. Problem description
Measurements and observed phenomena are important for the creation and validation of the
numerical model. Therefore, we used the values of the measured experimental data [6] input to the
simulation software and a comparison of the outputs. The purpose of these procedures is to obtain a
simulation model which can provide valid data to calculate energy consumption.
For modeling the problem, the specialized software in thermodynamic simulations ANSYS-Fluent
was used. The geometry and the mesh of the domain are shown in Figure 6. It was performed using
ANSYS Design Modeler and ANSYS Meshing program.
Exterior
Te =29.4oC
Interior
Ti =27.4oC
Fig. 2 - The mesh of the domain
Based on the experimental faade, in situ, and the results [9], simulations were performed under the
following assumptions and boundary conditions:
-
Faade geometry: H = 2.8 m, L = 2.0 m;

Solar radiation channel with constant heat flux: Qr = 248 W/m2, spectral factor 0.5;
Air temperature at the admission zone: ti = 29.4 C;
Room temperature: 27.4 C;
Air velocity at the entrance to the channel: vi = 0.51 m / s;
Characteristics of glazing thickness: 6 mm, absorption, reflection and transmission
coefficients are: = 0.11, = 0.07, = 0.82;
Ventilation system: exterior-exterior (air circulating upward channel between input-output
sections);
Flow regime: turbulent.
2.2. Numerical results

Numerical results have been materialized in the form of temperature and velocity spectrum and
profiles. The values obtained are used in RETScreen software to provide energetic and economic
information.
79
Velocity (m/s)
Temperature (C)
a)
b)
Fig. 3 - Temperatures (a) and velocities (b) spectrum
The temperature of the inner glass is reaching a maximum of 46 C (fig. 3.a.), and some energy
that is accumulated in the glazing must be evacuated. This is realized with good efficiency, in
this configuration, by the supply air that is circulating near it (fig. 3.b.), with temperatures
between 29 C and 32 C.
a)
b)
Fig. 4 - Profiles of temperature (a) and velocity (b) in median zone at different height of the channel
Temperature (C)
Fig. 5 - The numerical results comparing to experimental smoke visualizations [6]
In fig. 4 and 5 we can observe the temperature variation on the surfaces and inside the channel with
greater values on the exterior glass and smaller ones on the interior glass. The air flow dynamics in
the channel (fig. 4 and 5) shows a decreasing velocity near the interior surface with the height and a
recirculation zone near the exterior surface of the channel.
80
The qualitative comparison between experimental [6] and numerical results show a good agreement
which validates the last ones. Also, in table 1 are summarized the quantitative experimental [6] and
numerical results for the studied case, with a good accordance. The CFD results are confirmed by the
experimental ones with differences of 1522%. In this way, the numerical results were used as input
values for energetic calculation, such as the average temperatures of surfaces and of the air inside the
channel, in order to obtain the thermal balance of the studied room.
Table 1
Experimental and numerical results
Experimental results [6]
Numerical results
Differences
36,8 oC
0,21 m/s
0,62 m/s
31,3 oC
0,163 m/s
0,52 m/s
15%
22%
16%
Tmed,channel
vair,h=1,4m
vair,output
3. Energetic and economic analysis

The purpose of the study is to determine the improvement of the performance of the building by
realizing a comparative study of the energy consumptions for two configurations. The first
configuration, named base case, is the room without double skin faade and the second one, the
proposed case, consists of the same room equipped with the double skin faade that has the same
dimensions as the experimental model [6].
Numerical results show that the balance between costs and savings over the life of the project
represents an attractive proposition in financial terms.
3.1. Sheet 1: Energetic model
Fig. 6 - Energy consumption generated by RETScreen
In Fig. 6 a diminishing of energy consumption and a decrease in fuel consumption after

implementing the proposed case is observed. The values for necessary heating and cooling energy for
base case and proposed case were calculated using RETScreen internal option. The difference of
energy consumptions for heating (from 19 GJ to 15 GJ) and cooling (from 5 GJ to 4 GJ) are
determined by the superior insulation of the room equipped with DSF channel compared to the same
space with single glazing.
81
3.2. Sheet 2: Analysis of greenhouse gas
Fig. 7 - Summary for reduction of greenhouse gas emissions
By using double-skin faades, the environmental impact of greenhouse gas emissions is significantly
reduced (fig. 7) resulting in a net reduction in GHG emissions of 0.8 t CO2/year and 40t CO2/50
years respectively (the lifetime of the project).
3.3. Sheet 3: financial Analysis
Fig. 8 - Analysis of the financial viability of the project
Fig. 8 shows a number of parameters that highlight the viability of the analyzed case: Internal rate of
return > 5%, positive net present value and Cost-benefit ratio > 1.
Fig. 9 - Depreciation and getting profit on investment
After implementing the proposed case, the return on investment takes a very short time: between 5
and 6 years.
82
4. Conclusions
According to the analysis conducted in this paper, the systems consisting of ventilated doubleskin faades are certainly a solution to reduce the energy consumption of a building. The
simulations were done for a room with double skin faade similar to [6]. The numerical results
were validated by the experimental ones, in order to make an analysis on energy balance
between costs and savings over the life of the project.
Even though the initial investment is bigger that a building without double skin facade, it is
noted that implementing this type of faade is profitable on the medium and long term. The
return on investment takes place in a relatively short period of time. Reducing energy
consumption is both directly, by creating greater thermal insulation, and also indirectly, for
example if the faade is coupled to the ventilation and air conditioning system.
Viability of the project is highlighted. In the summer the role of DSF is to minimize solar loads,
using the protections and the ventilation of the channel. In winter it may be used as a seal
between the exterior and interior channel or to provide for rooms fresh air in the HVAC system
to a temperature higher than that outside temperature. This preheating outdoor air is done in
small part due to incident solar radiation. According to the results obtained with thermal
simulation software, a large amount of energy is recovered from the interior glazing that is
normally wasted energy.
References
[1] Saelens D. - Energy performance assessment of single storey multiple-skin faades, Katholieke Universiteit
Leuven, Faculteit Toegepaste Wetenschappen, Departement Burgerlijke Bouwkunde, Laboratorium
Bouwfysica, Kasteelpark Arenberg 51, B-3001 Leuven.
[2] Dickson A. - Modelling Double-Skin Faades, Department of Mechanical Engineering University of
Strathclyde, Glasgow UK.
[3] Safer N., 2006 - Modlisation des faades de type double-peau quipes de protections solaires: Approches
multi-chelles, LInstitut National des Sciences Appliques de Lyon.
[4] Gavan V., 2009 - Full-scale experimental evaluation and modeling of a double-skin faade. Optimal control of
thermal and visual comfort, LInstitut National des Sciences Appliques de Lyon.
[5] Poirazis H., 2004 - Double Skin Faades for Office Buildings - Literature Review, Division of Energy and
Building Design Department of Construction and Architecture, Lund Institute of Technology, Lund University,
Report EBD-R--04/3
[6] Chereche M, .a., 07.04.2011 - Comportarea termo-higro-energetic a faadelor ventilate, aplicabile la
cldirile publice, n condiiile climatice din Romnia, Research Report MDRT, Contract no. 493 /
[7] S. Hudisteanu, N.-C. Chereches, M. Chereches, 2012 - Numerical modeling of solar radiation inside ventilated
double-skin faade, Mathematical Modelling in Civil Engineering Journal, Nr.4, Vol. 8, pg. 96-101
[8] ANSYS Fluent - Documentation.
[9] Contract No. 493/07.04.2011 Ministry of Regional Development and Public Administration National Institute
for Research and Development in Construction, Urban Planning and Sustainable Spatial Development, NR&DI
URBAN-INCERC Iai Branch Thermal, hygrothermal and energetic comportment of ventilated facade, for
public buildings, for Romania climatic conditions. Pre-normative research.
83
OPEN SOURCE 3D MODELING FROM RASTER IMAGES

Gabriel Adrian KEREKES Master's student , Technical University of Civil Engineering, Faculty of Geodesy, email: kerekes.gabriel.adrian@gmail.com
Abstract: Three-dimensional models are important for civil engineering and architecture when
speaking of new projects. But if a 3D model of existing objectives is needed, laser scanners are to
be considered. Nevertheless, the process of data acquisition is time consuming and expensive.
This paper presents an alternative solution that involves 3D modeling using raster images acquired
with a simple digital camera. The principle of triangulation is applied for each pixel, after which
these are matched corresponding to their color and position. Moreover, the camera position is
generated by resection. The resulting model is made out of point clouds and polygonal surfaces that
can be edited and processed in order to obtain a realistic representation of that certain objective.
All of the mentioned above issues can be achieved using Open Source software in a simple, quick
and effective manner.
Keywords: Matching, calibration, alternative, triangulation, pixel.
1. Introduction
Considering the high development of data acquisition and processing, it can be inferred that
sometime soon there will be a digital replica of the real world. This can be observed in any
environment that uses a Geographic Information System (GIS), because two-dimensional
attribute data are not enough anymore and vertical development along with the complex
analysis that is required are a fact of today's world. A major role in this direction is attributed
to the three-dimensional (3D) models of artificial and natural elements. Nevertheless,
producing such models requires many resources and very often the process is time consuming.
Visualization of 3D models is important in many fields because they are a true representation
and simplify decision making.
This paper is presenting an alternative solution that contributes to the production of 3D models
for buildings or other structures from raster images. Nowadays, there are many hardware and
software solutions from the surveying field that provide cloud points for different objects,
worth mentioning being the 3D laser scanner (terrestrial or airborne), but the costs of such
instruments and programs are very high. An alternative solution is to use Open Source
software, freely available, along with passive sensors for capturing data, category to which
digital cameras belong.
2. Data acquisition
Usually for 3D laser scanning a dedicated laser scanner is recommended, but a cloud point can
also be generated by other means. It can be said that this method is somewhere in between
photogrammetry and laser scanning, combining photogrammetric algorithms with 3D modeling
techniques used in laser scanning. From these details, only some that are important will be
presented to help in understanding the basic principles of 3D reconstruction from images.
The Open Source program used initially is called Insight 3D and it is downloadable for free at
the following address: http://insight3d.sourceforge.net/. At the same address some tutorials and
sample images are available so that the user can get a quick idea of how the program works. This
program allows 3D modeling from raster images using the principle of triangulation and
resection. To highlight the program's capacity some sample images captured with a Nikon S8000
camera were used and as objects of the case study some buildings from Bucharest and Arad were
chosen. Reconstructing the reality thorough 3D models from images implies capturing the whole
object; this means that photos should be taken so that there are no shadowing areas. Pictures can
84
be taken from any angle without measuring or using a tripod. The program can determine the
position of the camera with algorithms similar to those used in resection. However, for good
results, there are a series of recommendations: images must be focused blurred images are of
no use, large overlaps between neighboring photos are preferred so that the coincidence of
some areas can be established, the limited angle difference should be somewhere between 15
25 - acute angles give good results, details and textured models are preferred because no details
(like modern faades) give no information to the program, planar scenes are to be avoided, image
cropping is not allowed so the resolution should be uniform and unnecessary parts can be
deleted after the algorithms have been applied and finally, each part that needs reconstruction
should be visible in at least three scenes also known as the rule of three. [1] An example that
respects the mentioned rules can be seen in figure 1.
Fig. 1 Series of captured images of the Architects Order building that comply with the above mentioned principles
3. Data processing
Once the images were captured, the processing can be initiated. These photos will be saved so
that they can be used in another program that uses similar principles, but works as an online
service. After adding the photos in Insight, they appear in the work space and can be visually
analyzed by the user. In the next step, the program will identify common points from the loaded
images, phase that is considered the most important one. The problem of correspondence is now
raised. [2] This implies identifying corresponding areas from two images of the same object but
taken from different angles. To solve this, matching techniques are used for the common pixels
of two images analyzed as a series.
The user can select the scanning resolution of the image,
thus determining the number of details that the resulting
model will contain. Depending on the number of images
that are loaded and the processor power, the matching
phase is more or less time consuming. Nevertheless, with
seven images and a high resolution setting, a cloud point
represented in figure 2 is obtained within 15 minutes. At
this stage the user can manually select vertices (points in
space) if these are not automatically identified. It can be
observed that the program works best when there are lots
of details, thus
the lower part
of the building
will
be
represented with a greater degree of details (the old
faade), whereas the superior part having the same
color was not properly described.
Fig. 2 Automatic point matching
The next step implies identifying the cameras

positions also referred to as the points from which the
shots were taken and the calibration parameters.
Fig. 3 The cameras positions
85
Knowing the image coordinates in a local system, the program determines the camera
coordinates by resection. (Figure 3)
At this moment the common points and the camera positions and parameters are known for each
image, therefore the triangulation of the matched points can be commenced. Worth mentioning is
that not all the matched points are required for the final model so the user can select to
triangulate only the resulted points filtered by the RANSAC method (Random Sample
Consensus). This algorithm is used in image processing and computer vision to eliminate the
outliers from a set of measurements. (Figure 4) It permits the estimation of mathematical model
that closely resembles reality through an iterative process. [3]
If the user desires to manually add more
points, this is facilitated by Insight 3D
through interrogation windows of the
found points. Black boxes resemble all the
common points and green points are
representative for the deviations from the
automatically identified position of that
specific point. These should be in
coincidence in an idealistic situation. The
more distant the green point is from the
Fig. 4 The mathematic model for RANSAC filtering
black box the greater the errors. If a green

point is above the corresponding black
box its deviation is small; to see this a
user can click on that point and observe
that the same corner of the building has
been identified. (Figure 5)
The next step implies deleting all the
unnecessary vertices by selecting them (in
a similar manner with other CAD
software) and repeating this operation for
each image, leaving only points on the
object of interest. The user is still allowed
Fig. 5 Triangulated points automatically identified.
at this stage to add points if he desires.
Triangulation is initiated once more and if the mentioned above conditions have been met with
enough points identified, a satisfactory model is produced. The preliminary result will be a cloud
point with all the determined elements. However, there are some objects with aberrant positions;
these include exterior objects like trees, cars or points on the sky, which will be manually
eliminated because they can badly influence the final model. To easily identify the elements
from the image and the cloud points, the vertices can be colorized with their corresponding pixel
color. (Figure 6) Beside the generated cloud point, the program permits polygonal surface
creation and texture generation for these
Fig. 6 The obtained cloud point
polygons. If the shapes resemble simple
geometric volumes, they can be captured as
polygons, but if there are curved surfaces, this
becomes useless. All of these steps can be more
or less troublesome depending on the
complexity of the reconstructed model.
Even though the presented algorithm leads us to
a raw 3D model, this cannot comply with all the
demands sometimes because of the low point
density. It is just not enough for all purposes,
but for post processing the cloud point can be
86
exported as an ASCII file. [4] Another factor that should be taken into consideration is the
processor power of the computer and the video boards, many projects requiring powerful hardware
configuration and resources that the simple user does not possess. Being an Open Source program
implies that certain bugs and errors are not verified so the stability of the program is sometimes
unreliable. These were a few downfalls of Insight 3D.
There are other Open Source solutions that can be used to refine the created model with Insight 3D,
but in this paper another option of obtaining a cloud point was preferred. It implies the same
principles as Insight 3D reconstructing 3D models from cloud points - but works as an online
service called My3Dscanner. In this case, the user uploads a set of images that corresponds to the
conditions mentioned before and through a dedicated server, these will be processed and the user
will receive free of any costs a dense cloud point resembling the captured object. This is a huge
advantage because density means greater details and a high similarity with the real objective.
(Figure 7) Another great thing is that the number of images that can be uploaded is greater than
those that Insight 3D supports.
For data visualization and processing in the cloud point format there are many Open Source
programs out of which two where used because of their efficiency, namely: Meshlab and
CloudCompare. This allows the reconstruction of a 3D model from a set of raw data and other
operations and statistics. The strong point of this method is that details can be analyzed in a short
time due to the high cloud point density. After loading the cloud points, an approximation for the
surface is made using the Delaunay triangulation algorithm. This new generated surface is colorized
through the attribute transfer function provided by Meshlab, thus the attributes of the points are
transferred to the created mesh. Many cloud point files contain information like spatial coordinates
(X, Y, Z), RGB values and normals on three axes. Using this kind of info the mesh is colorized and
the resulting model nicely resembles reality after only some quick refinements. (Figure 8)
Fig. 7 Cloud point generated by My3dscanner
Fig. 8 Mesh created by Delaunay triangulation and

attribute transfer
Fig. 9 3D model of the Romanian Athenaeum Bucharest
87
Other case studies are presented because they are representative for the purpose of this paper,
mainly creating 3D models from raster images. Another fact is that the chosen buildings do not
contain geometric volumes that could be easily recreated through classical measurements.
Fig. 10 Ioan Slavici Theatre - Arad
Fig. 11 Catholic Cathedral - Arad
Fig. 12 Old house from the historic center - Arad
From the created models one can see that the building details are important for the 3D model
reconstruction and that a building with few details is hard to reconstruct because of the
intricacies of identifying each pixel. A disadvantage is that for the shadowed areas (not captured
within the image) the model will contain holes. To solve this problem, pictures can be taken
from different heights. The elements for each created model are presented in table 1. This helps
at getting an overview of the necessities implied in the construction and processing of any 3D
model created by these means.
Table 1
Elements necessary for 3D models reconstruction
Building
Architects Order
Romanian Athenaeum
Ioan Slavici Theatre - Arad
Old house - Arad
88
Images
7
9
18
15
Points (cloud points)

106481
243464
315379
346739
Triangles (Faces)
197053
485643
628858
572478
4. Conclusions
After the whole process one can affirm that through simple operations and free software 3D
models from raster images can be obtained, which can serve many purposes. Some practical
applications of this method can be found in fields like: architecture, archeology, civil
engineering, geology, geodesy, photogrammetry, surveying in engineering etc. Another domain
that could benefit from this is the latest techniques of 3D printing. 3D models resembling realscale objects can now be printed using the latest 3D printers, and to do this one needs a digital
model of that certain object.
Other useful applications could be encountered in the monitoring and deformation analysis of
different natural and artificial object or phenomenon. [5] For example, cloud points can be
reconstructed from different sessions of measurements and then compared. In what concerns
earth works, a quarry or sand pit could be modeled in different sessions and then determine the
volume of material that has been extracted through a simple comparison of two sets of cloud
points. Of course to have real units, the model should be brought to scale and georeferenced.
Similarly, in monitoring buildings a deformation analysis can be performed through a cloud
comparison. The ideal Open Source program for this is Cloud Compare.
Further research is needed to verify the precision of such models and to establish a relation
between accuracy and the used instruments for data acquisition.
References
[1]. ***Insight 3D Tutorial - http://insight3d.sourceforge.net/
[2]. Abhijit S. Ogale, Yiannis Aloimonos - Shape and the stereo correspondence problem, in International Journal
of Computer Vision, Vol. 65, No. 1, Oct. 2005.
[3]. Martin A. Fischler and Robert C. Bolles - Random Sample Consensus: A Paradigm for Model Fitting with
Applications to Image Analysis and Automated Cartography, in Communications of the ACM Vol. 24, No. 6,
Iun 1981, pp. 381395.
[4]. Coarc, C. Sisteme de msurare n industrie, Editura Conspress, Bucureti, 2009
[5]. Katharina Bruhm, Manfred Buchroithner , Bernd Hetze - True-3D Visualization of Glacier Retreat in the
Dachstein Massif, Austria: Cross-Media Hard- and Softcopy Displays, Developments in 3D Geo-Information
Sciences, in Lecture Notes in Geoinformation and Cartography, Editura Springer, 2010
89
INFLUENCES OF OXIDATION STEP AND INITIAL METAL

CONCENTRATIONS ON IRON AND MANGANESE
REMOVAL EFFICIENCY
Alexandru JERCAN PhD Student, Department of Hydro-technical Engineering, Faculty of Hydrotechnics,
Technical University of Civil Engineering of Bucharest (UTCB), e-mail: alexandru.jercan@gmail.com
Abstract: Iron and Manganese can be found at varying concentration levels in both surface and
ground waters. Even at low concentrations, both metals can cause numerous drinking water
problems like turbidity, discoloration and bad taste. Given the aesthetic and consumer acceptability
issues, removal of iron and manganese is an important aspect in selecting the best water treatment
plant design for a given water source.
The paper presents the results obtained in iron and manganese removal tests conducted on synthetic
water samples. The tested processes were oxidation with air and potassium permanganate, followed
by filtration, for different iron and manganese initial concentrations.
The results of the experimental tests revealed that the removal efficiency is strongly influenced by
oxidation conditions and initial iron and manganese concentrations.
Keywords: water treatment, chemical oxidation, water filtration, drinking water acceptability.
1. Iron and manganese contribution to drinking water problems

Iron and manganese are relatively soluble under reducing conditions, for example, in
groundwaters, stagnant surface waters, and in the hypolimnetic waters of eutrophic lakes,
reservoirs, and impoundments. Correspondingly, they are quite insoluble under oxidizing
conditions, for example, in flowing streams and in the epilimnetic waters of lakes or
impoundments, or in hypolimnetic waters that have been subject to hypolimnetic aeration to
maintain oxidizing conditions. The reduced forms of iron and manganese-ferrous iron, Fe(II),
and manganous manganese, Mn(II)-may occur as the free metal ions, Fe2+ and Mn2+, which is
often the case in most groundwaters, or they may be found complexed to various degrees with
natural organic matter, as is often the case in surface waters and highly colored groundwaters.
During and immediately following lake overturn- that is, when the iron- and manganese-rich
hypo limnetic water is mixed with the remainder of the lake water-dissolved iron and manganese
levels in the upper portions of the lake can increase appreciably [1].
The primary concern with elevated levels of dissolved iron and manganese in water is that when
they become oxidized to insoluble ferric hydroxide, Fe(OH)3(s), and manganese dioxide,
MnO2(s), they precipitate and cause reddish-orange or black deposits, respectively, to appear on
plumbing fixtures and to create stains during laundering operations [1].
Anaerobic groundwater may contain ferrous iron at concentrations of up to several mg Fe/l
without discoloration or turbidity in the water when directly pumped from a well. On exposure to
the atmosphere, however, the ferrous iron oxidizes to ferric iron, giving an objectionable
reddish-brown color to the water. Iron also promotes the growth of iron bacteria, which derive
their energy from the oxidation of ferrous iron to ferric iron and in the process deposit a slimy
coating on the piping. At levels above 0.3 mg Fe/l, iron stains laundry and plumbing fixtures.
There is usually no noticeable taste at iron concentrations below 0.3 mg Fe/l, although turbidity
and color may develop [2].
At levels exceeding 0.1 mg Mn/l, manganese in water supplies causes an undesirable taste in
beverages and stains sanitary ware and laundry. The presence of manganese in drinking water,
like that of iron, may lead to the accumulation of deposits in the distribution system [2].
90
Given the drinking water consumer acceptability problems generated by these elements,
reducing their concentrations is an important objective of the treatment process for any drinking
water treatment plant that relies on raw water rich in soluble iron and manganese ions.
The most common iron and manganese removal technologies are based on soluble ions oxidation
and separation of the resulting insoluble ions by clarification and/or filtration.
The goal of the study presented in this paper is to evaluate the influence of oxidation step and
initial soluble ions concentrations on iron and manganese removal efficiency.
2. Iron and manganese oxidation tests and experimental results
Iron and manganese oxidation tests were conducted on synthetic solutions for various initial
soluble ions concentrations, for both single type ions solutions and both type ions solutions.
The oxidation agents used in the tests were atmospheric oxygen and potassium permanganate,
for the following experimental conditions:
-
Temperature: 20 C;
Air flow: 5 l/min;
Potassium permanganate stoichiometric dose KMnO4:Mn= 1.92 mg KMnO4/mg Mn;
Potassium permanganate stoichiometric dose KMnO4:Fe= 0.94 mg KMnO4/mg Fe;
Samples were collected at the beginning of each test and after 10, 20, 40 minutes respectively.
Each synthetic solution was labeled by the initial iron, manganese and iron+manganese
concentrations and it was oxidized by oxygen and/or potassium permanganate. Every collected
sample was passed through membrane filters and the residual iron and/or manganese
concentrations were measured for the filtered water.
2.1. Experimental results for oxidation by air-iron removal tests
The following table presents the results obtained in the iron removal tests using oxidation by air.
Table 1
Residual iron concentrations obtained in iron removal tests using oxidation by air
Solution label
Fe2+ 1mg/l
Fe2+ 0.5 mg/l
Fe2+ 0.25 mg/l
Water samples collected before oxidation
Water samples collected at air oxidation time:
Water samples before

filtration
Water samples after

filtration
10 min
20 min
40 min
Fe=0.65 mg/l
Fe= 0.15 mg/l
Fe= 0.15 mg/l
Fe= 0.15 mg/l
pH=6.49
pH=7.50
pH=7.70
pH=7.70
NTU=0.76
NTU=2.72
NTU=2.99
NTU=3.01
Fe= 0.29 mg/l
Fe= 0.10 mg/l
Fe= 0.07 mg/l
Fe= 0.05 mg/l
pH=6.86
pH=6.95
pH=7.68
pH=7.93
NTU=0.99
NTU=1.72
NTU=1.74
NTU=1.75
Fe= 0.03 mg/l
Fe= 0.02 mg/l
Fe= 0.02 mg/l
Fe= 0.01 mg/l
pH=7.15
pH=7.20
pH=7.68
pH=7.91
NTU=1.14
NTU=1.20
NTU=1.27
NTU=1.41
Fe= 0.95 mg/l
Fe= 0.51 mg/l
Fe= 0.29 mg/l
91
2.2. Experimental results for oxidation by potassium permanganate-manganese removal

tests
The following table presents the results obtained in the manganese removal tests using oxidation
by potassium permanganate.
Table 2
Residual manganese concentrations obtained in manganese removal tests using oxidation by potassium
permanganate
Solution label
Mn2+ 1mg/l
Mn2+ 0.5 mg/l
Mn2+ 0.25 mg/l
Filtered water sample collected

before oxidation
Filtered water samples collected at oxidation time:

10 min
20 min
40 min
Mn=0.332 mg/l
Mn= 0.049 mg/l
Mn= 0.019 mg/l
Mn= 0.008 mg/l
NTU=0.97
NTU=1.68
NTU=2.21
NTU=2.99
Mn=0.196 mg/l
Mn= 0.138 mg/l
Mn= 0.112 mg/l
Mn= 0.108 mg/l
NTU=1.01
NTU=1.42
NTU=1.64
NTU=1.81
Mn=0.128 mg/l
Mn= 0.088 mg/l
Mn= 0.046 mg/l
Mn= 0.022 mg/l
NTU=1.05
NTU=1.23
NTU=1.36
NTU=1.48
2.3. Experimental results for oxidation by air and potassium permanganate-iron and
manganese removal tests
The following tables present the results obtained in the iron and manganese removal tests using
oxidation by air and potassium permanganate.
Table 3
Residual iron and manganese concentrations obtained in iron (1 mg Fe/l) and manganese (0.3 mg Mn/l)
removal tests using oxidation by air and potassium permanganate
Initial iron
concentration
Fe2+ 1.0 mg/l
Initial manganese
concentration
Mn2+0.30 mg/l
Filtered water
sample collected
before oxidation
Filtered water sample

collected at oxidation
time:
10 min

time:
20 min

time:
40 min
Fe* = 0.87 mg/l
Fe* = 0.01 mg/l
Fe* = 0.00 mg/l
Fe* = 0.00 mg/l
Fe = 0.83 mg/l
Fe = 0.01 mg/l
Fe = 0.00 mg/l
Fe = 0.00 mg/l
pH=6.41
pH=6.64
pH=6.93
pH=7.52
Mn*=0.264 mg/l
Mn*=0.364 mg/l
Mn*=0.321 mg/l
Mn*=0.335 mg/l
Mn=0.322 mg/l
Mn=0.044 mg/l
Mn=0.047 mg/l
Mn=0.051 mg/l
*-results obtained for a test conducted with KMnO4 under dosing.

Table 4
Residual iron and manganese concentrations obtained in iron (0.5 mg Fe/l) and manganese (0.15 mg Mn/l)
Initial iron
concentration
Fe2+ 0.5 mg/l
Initial manganese
concentration
Mn2+0.15 mg/l
92
Filtered water
sample collected
before oxidation

time:
10 min

time:
20 min

time:
40 min
Fe = 0.46 mg/l
Fe = 0.01 mg/l
Fe = 0.01 mg/l
Fe = 0.01 mg/l
pH=7.40
pH=7.55
pH=7.80
pH=7.99
Mn=0.111 mg/l
Mn=0.038 mg/l
Mn=0.020 mg/l
Mn=0.015 mg/l
Table 5
Residual iron and manganese concentrations obtained in iron (0.25 mg Fe/l) and manganese (0.075 mg Mn/l)
Fe = 0.18 mg/l
pH=7.20

time:
10 min
Fe = 0.02 mg/l
pH=7.52

time:
20 min
Fe = 0.01 mg/l
pH=7.84

time:
40 min
Fe = 0.01 mg/l
pH=7.92
Mn=0.087 mg/l
Mn=0.044 mg/l
Mn=0.032 mg/l
Mn=0.031 mg/l
Filtered water
sample collected
before oxidation
Initial iron
concentration
Fe2+ 0.25 mg/l
Initial manganese
concentration
Mn2+0.075 mg/l
3. Experimental results interpretation

3.1. Iron removal tests - Oxidation by air experimental results interpretation
Figure 1 presents residual iron concentrations evolution in time. The experimental results lead to
the following conclusions:
-
For each test, the most part of the iron was removed by filtration (membrane filtration
was used); as the graphic shows, for the initial moment two iron concentrations were
registered, one before and one after filtration;
After the first 10 minutes of each test, the iron oxidation rate registered a slight decrease;
The highest iron oxidation rate occurred during the first 10 minutes of each test.
1.00
1.20
Fe1mg/l
0.90
Fe0.25mg/l
Fe(mg/l)
0.70
0.60
0.50
0.40
0.30
0.20
Ironconcentration(mg/l)
Fe0.5mg/l
0.80
Initialironconcentration[mg/l]
Residualironconcentration[mg/l]
1.00
0.80
0.60
0.40
0.20
0.10
0.00
0.00
0
10
20
30
40
1.00
Oxidationtime(min)
Fig. 1 Residual iron concentration evolution in

time.
0.50
FeTests1,2,3
0.25
Fig. 2 Initial/residual iron concentration comparison.
Figure 2 presents comparisons between the initial and the residual iron concentrations.
As shown in the graphic, the highest reduction degree was registered for the highest initial iron
concentration (1 mg Fe/l).
Figure 3 presents water sample turbidity evolution in time for the tested synthetic solutions. The
graphic reveals the following:
-
For all tests, water turbidity increased during the oxidation process;
The highest turbidity value at the end of the experimental tests was registered for the
solution having the highest initial iron concentration (1 mg Fe/l).
93
4.00
Fe1mg/l
Fe0.5mg/l
Fe0.25mg/l
Turbidity(NTU)
3.50
3.00
2.50
2.00
1.50
1.00
0.50
0
10
20
30
Oxidationtime(min)
40
Fig. 3 Water turbidity evolution in time.
Fig. 4 Iron removal efficiency evolution
Figure 4 presents iron removal efficiency evolution in time for the tested synthetic solutions. The
graphic reveals the following:
-
For the solution having the highest initial iron concentration (1 mg Fe/l), iron removal
efficiency registered an important increase in the first 10 minutes of air oxidation, and
after that it maintained a relatively constant evolution;
For the solution having an initial iron concentration of 0.5 mg Fe/l, iron removal
efficiency also registered an important increase in the first 10 minutes of air oxidation,
and after that it kept increasing upon reaching a value of about 80%;
efficiency registered a relatively linear evolution;
A general conclusion is that iron removal efficiency is higher for solutions with higher
initial iron concentration.
3.2. Manganese removal tests - Oxidation by potassium permanganate experimental results

interpretation
The experimental results lead to the following

conclusions:
- The highest manganese oxidation rate
occurred during the first 10 minutes of
each test;
- For solutions with lower initial manganese
concentrations (0.2 mg Mn/l and 0.1 mg
Mn/l) the registered reduction rate is
almost linear for the first 20 minutes of the
oxidation test;
- After the first 20 minutes of each test,
manganese oxidation rate registered a
slight decrease.
94
0.350
Mn0.3mg/l
Mn0.2mg/l
0.300
0.250
Mn (mg/l)
Figure 5 presents residual manganese

concentrations evolution in time.
0.200
0.150
0.100
0.050
0.000
0
10
20
30
Oxidationtime(min)
40
Fig. 5 Residual manganese concentration evolution

in time
Figure 7 presents water sample turbidity

evolution in time for the tested synthetic
solutions.
The graphic reveals the following:
- For all tests, water turbidity increased
during the oxidation process;
- The highest turbidity value at the end of
the experimental tests was registered for
the solution having the highest initial
manganese concentration (0.3 mg Mn/l).
0.350
0.200
0.150
0.100
0.050
0.000
0.300
Turbidity[NTU]
100
Eficien(%)
Mn0.3mg/l
Mn0.2mg/l
Mn0.1mg/l
1.00
0.50
0
10
20
30
Oxidationtime(min)
40
Fig. 7 Water turbidity evolution in time.
0.100
Fig. 6 Initial/residual manganese concentration

comparison
3.00
1.50
0.200
MnTests1,2,3
120
2.00
Residualconcentration[mg/l]
0.250
3.50
2.50
Initialconcentration[mg/l]
0.300
Mn concentration (mg/l)
Figure 6 presents a comparison between

initial
and
residual
manganese
concentrations. As shown in the graphic, the
highest reduction degree was registered for
the highest initial manganese concentration
(0.3 mg Mn/l).
80
60
40
Mn0.3mg/l
Mn0.2mg/l
Mn0.1mg/l
20
0
0
10
20
30
Oxidationtime(min)
40
Figure 8 presents manganese removal efficiency evolution in time for the tested synthetic
solutions. The graphic reveals the following:
-
For the solution having the highest initial manganese concentration (0.3 mg/l),
manganese removal efficiency registered an important increase in the first 10 minutes of
oxidation, and after that it maintained a relatively constant evolution;
For the solutions having initial manganese concentrations of 0.2 mg Mn/l and 0.1 mg
Mn/l, respectively, manganese removal efficiency also registered an important increase,
lower than for the solution with 0.3 mg Mn/l, in the first 10 minutes of oxidation. After
the first 10 minutes of oxidation removal efficiency increased for the solution with 0.2
mg Mn/l up to 80%, and for the solution with 0.1 mg Mn/l up to 40%;
For the solution containing an initial manganese concentration of 0.3 mg Mn/l, almost
100% removal efficiency was registered;
A general conclusion is that manganese removal efficiency is higher for solutions with
higher initial manganese concentration.
95
3.3. Iron and Manganese removal tests - Oxidation by air and potassium permanganate
experimental results interpretation
During the first 10 minutes residual iron

concentrations was reduced significantly;
At the end of the test iron reduction

efficiency was 100%;
Under dosing KMnO4 resulted in no

manganese reduction, as at the end of the
test residual Mn concentration was in
fact higher than the initial concentration.
1.00
0.90
Fe1mg/l
0.80
Mn0.3mg/l
0.70
Fe,Mn (mg/l)
Figure 9 presents residual iron and

manganese concentrations evolution in time
for a test conducted with oxidation by air and
KMnO4 under dosing. The experimental
results lead to the following conclusions:
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0
10
20
30
Oxidationtime(min)
40
Fig. 9 Residual iron and manganese concentration

evolution in time for KMnO4 under dosing test
Figure 10 presents the residual iron concentrations evolution in time. The experimental results
lead to the following conclusions:
- For each of the initial iron concentration solutions, the most part of the iron was removed
during the first 10 minutes of the tests;
- Up to the end of each test, iron reduction efficiency was close to 100%;
- Iron removal efficiency for the oxidation tests with air and KMnO4 was higher than in
the case of tests where oxidation was realized only by air;
Figure 11 presents residual manganese concentrations evolution in time. The experimental
results lead to the following conclusions:
- For the solution having an initial concentration of 0.3 mg Mn/l, most manganese
reduction was completed in the first 10 minutes, after which reduction rate decreased
significantly;
- For the solution having initial concentrations of 0.2 mg Mn/l and 0.1 mg Mn/l,
respectively, manganese reduction rate was almost linear for the first 20 minutes of each
test, after which it registered a slight decrease;
- Manganese removal efficiency for oxidation tests with air and KMnO4 was higher than
in the case of tests where oxidation was realized only with KMnO4.
0.350
0.90
0.80
0.300
Fe0.5mg/l
0.70
0.250
Fe0.25mg/l
Mn (mg/l)
0.60
Fe(mg/l)
Mn0.3mg/l
Mn0.15mg/l
Mn0.075mg/l
Fe1mg/l
0.200
0.50
0.150
0.40
0.30
0.100
0.20
0.050
0.10
0.000
0.00
0
10
20
30
Oxidationtime(min)
40
Fig. 10 Residual iron concentration evolution in time.
96
10
20
30
Oxidationtime(min)
40
Fig. 11 Residual manganese concentration evolution in

time.
Figure 12 presents comparisons between the

initial and the residual iron concentrations. As
shown in the graphic, for the highest initial
iron concentration (1 mg Fe/l), the iron
reduction efficiency was practically 100%.
Figure 13 presents comparisons between the

initial
and
the
residual
manganese
concentrations.
As shown in the graphic, all tests resulted in
residual manganese concentrations encircling
the limit established by law.
Fig. 12 Initial/residual iron concentration comparison.
Fig. 13 Initial/residual manganese concentration

comparison.
Figures 14 and 15 present iron and manganese removal efficiency evolution in time for the tested
synthetic solutions. The graphics reveal the following:
-
For all tests, an important increase was registered in the first 10 minutes of oxidation;
For the solutions having initial iron concentrations of 1.0 mg Fe/l and 0.5 mg Fe/l, iron
removal efficiency reached 100% in the first 10 minutes of oxidation;
efficiency was lower than for the other solutions with higher initial iron concentrations;
Iron removal efficiency for all oxidation tests with air and KMnO4 was higher than in the
case of tests where oxidation was realized only by air;
For the solutions having initial manganese concentrations of 0.15 mg Mn/l and 0.075 mg
Mn/l, manganese removal efficiency increased gradually between the samples collected
at 10 minutes interval, reaching values of up to 90% and 60%, respectively;
A general conclusion is that manganese removal efficiency is highly influenced by the
initial manganese concentration;
Manganese removal efficiency for all oxidation tests with air and KMnO4 was higher
than in the case of tests where oxidation was realized only with KMnO4;
120
100
90
100
80
60
Fe1mg/l
40
Fe0.5mg/l
20
Fe0.25mg/l
Efficiency (%)
Efficiency (%)
80
70
60
50
40
Mn0.3mg/l
30
20
Mn0.15mg/l
10
Mn0.075mg/l
0
0
10
20
30
Oxidationtime(min)
40
10
20
30
Oxidationtime(min)
40
Fig. 15 Manganese removal efficiency evolution
4. Conclusions
Iron and manganese represent important factors contributing to drinking water consumer
acceptability issues. In this regard, removal of iron and manganese is an important aspect in
selecting the best water treatment plant design for a given water source.
Iron and manganese oxidation tests were conducted on synthetic solutions for various initial
soluble ions concentrations, for single type ions solutions and mixed type ions solutions.
The tests confirmed that filtration has an important contribution in iron and manganese removal
processes. Also, the most intense part of the oxidation process occurred in the first 10 minutes of
each test.
For both iron and manganese, it was evidenced that higher initial concentrations lead to higher
removal efficiencies.
Manganese removal tests showed that for higher initial concentrations the oxidation rate is
higher in the beginning of the process, followed by oxidation rate decrease. It was observed that
for lower initial Mn concentrations the oxidation process develops at a more uniform rate.
The importance of correct oxidant dosing was evidenced, as in one of the tests KMnO4 under
dosing was used, resulting in oxidation of iron only, accompanied by an increase of soluble Mn
concentration.
By comparing the results of tests using single and mixed oxidant agents, it was observed that
very good removal speeds and efficiencies were achieved when using simultaneous oxidation by
air and potassium permanganate (residual concentrations of 0.000.01 mg Fe/l and less than
0.05 mg Mn/l were obtained).
As each raw water source has its own specific composition, choosing the best treatment process
requires thorough pilot-scale tests that guarantee best technical and economical results regarding
both oxidation step and iron and manganese removal step.
References
1.
2.
McGRAW-HILL, INC - Water Quality and Treatment a Handbook of Community Water Supplies, 1999
WHO - Guidelines for Drinking Water Quality, Third Edition, Volume 1, Geneva, 2006
Mathematical Modelling in Civil Engineering, Special Issue YRC November 2013
98
SEISMIC RESPONSE OF TALL BUILDINGS WITH ROCKING

WALLS SYSTEM
Lidia MARIN PhD Student, Technical University of Civil Engineering of Bucharest, lidia.mihaela@gmail.com
Mircea VADUVA PhD Student, Technical University of Civil Engineering of Bucharest,
e-mail: mircea.vaduva@gmail.com
Abstract: Rocking walls are an alternative solution to classical reinforced concrete walls. The
difference between them is that the first solution has the possibility to rock at the base. The main
purpose of this paper is to highlight the advantages and disadvantages of using the rocking wall
system for tall buildings. Four structures with different solutions for the concrete walls were
analyzed. For the first structure, classical reinforced concrete walls embedded at the base level were
used. In the second structure, the classical walls were replaced with rocking walls. For the third
structure, rocking walls were used with additional devices for seismic energy dissipation consisting of
milled segments. For the last structure, rocking walls and buckling restrained braces were used for
energy dissipation. In all four cases, static nonlinear analyses were conducted and the results were
compared in order to obtain an image about the seismic response of the rocking wall system.
Parameters like plastic rotations in structural elements, relative displacements between stories and
capacity curves were analyzed and compared to reach the purpose of this study. The results show that
the seismic response of tall buildings can be improved by using the rocking wall system.
Keywords: reinforced concrete walls, dissipative system, rocking walls, plastic deformation, tall
buildings
1. Introduction
The rocking wall system was proposed like an alternative to classical reinforced concrete walls
in order to achieve a better response to seismic motion. According to [1] the rocking walls are
fixed on the foundation throught unbounded post-tensioned tendons that are placed in the center
of the wall and remain in the elastic domain. The elastic force from the post-tensioned tendons is
the one that offers strength to lateral forces and gives the possibility to the system to come back
to the initial position.
In this study, the seismic response of a structure with 20 stories (64m height) that has different
solutions for the reinforced concrete walls was analyzed. For the first structure, classical
reinforced concrete walls embedded at the base level were used. In the second structure, the
classical walls were replaced with rocking walls. For the third structure, rocking walls with
additional devices for seismic energy dissipation consisting of milled segments were used. For
the last structure, rocking walls and buckling restrained braces for energy dissipation were used.
The purpose of this paper is to highlight the advantages and disadvantages of using rocking walls
system for tall buildings that are placed in seismic areas.
2. Building description
2.1. Geometry
In order to improve the computation speed and the dimensions of the information several
analysis on 2D models were made. In figure 1 are presented the four structures, which were
analyzed.
The dimensions of the structural elements are: beams 0.55x0.75m, columns 1.55x1.55m and
walls 0.50x7.0m. For a better understanding of the structural behavior of the rocking walls, three
solutions were analyzed. For the structure B just rocking walls without any improvement were
used.
99
b.
c.
d.
Fig. 1 a. Structure A - Structure with shear reinforced concrete walls embedded at the base level; b. Structure B Structure with rocking walls; c. Structure C - Structure with rocking walls and milled segments for seismic energy
dissipation; d. Structure D - Structure with rocking walls and buckling restrained braces for energy dissipation
Structure C has rocking walls with additional devices for seismic dissipation consisting of milled
segments. For the third solution that has been analyzed (structure D) the classical reinforced concrete
walls were replaced with rocking walls and buckling restrained braces for energy dissipation.
2.2. Loads
The main destination of the building is for offices and according to [2], the structure belongs to
category B, which means a live load of 2.5 KN/m2. The self-weight of the in plane elements was
automatic computed by the program and additional dead loads from the elements in transversal
section were modeled as distributed loads on the beams. The loads coming from partitioned
walls, from finishings and from technical areas were summarized in a 3.0 KN/m2 load. On the
roof a load of 4.5 KN/m2 was considered and the snow load was taken equal to 1.6 KN/m2.
2.3. Lateral load
The structural elements were designed based on equivalent static lateral forces method. The
seismic behavior factor is q=5*1.0=5.0 (specific to the dual structures). The seismic load was
computed according to [3] and the construction site had the following characteristics: peak
ground acceleration of 0.24g and a natural period of local soil of 1.6 seconds.
All the verifications were made using static nonlinear analysis. To determine the displacement
demand, the elastic response spectrum from [3] was used.
Structure modeling
2.4. Materials
For all the reinforced concrete elements concrete C30/37 was used and the steel for rebars was
BST500s (fyd=435N/mm2, ultimate strain capacity 0.10). The characteristics for the materials are
according to [4]. In figure 2 the properties of the materials used in this study are presented.
Fig. 2 - Properties of BST500s and concrete C30/37
The concrete elements were designed just for the structure A and for the others structures the
same elements were used. The difference between the four structures consists on the way the
100
Special Issue Y.R.C. - 2013
concrete walls are modeled. The material used for the tendons that were modeled in the structure
with rocking walls was picked up from [5]. The values for modulus of elasticity and yield
strength are 195GPa and fyd=1330N/mm2.The stiffness of the concrete used for the rocking walls
was computed using 100% of the modulus of elasticity, due to the fact that the walls are posttensioned and therefore the compression forces are big enough to keep the element uncracked.
2.5. Elements
The software used for this analysis was SAP2000. The columns, beams and walls were modeled
as frame elements. For a good approximation of the behavior of shear walls two types of
elements were used: mid-pear frames and very stiff beams which allow a good simulation of the
real behavior around the connection between the elements. In figure 3 is presented the model for
the concrete wall according to [6].
Fig.3 - The model of the structural wall [6]
For the structures with rocking walls the same materials, properties and cross sections of the
elements were used like in the first model. This time, the walls were modeled as shell elements and
the connections between them and the frames were considered rigid. The difference between the
first and the others models is that the rocking walls are not fixed on the base, they are placed on the
foundation. The interaction between the concrete walls and the foundations was modeled using gap
elements. The foundation is considered stiff enough to sustain the rocking of the wall without
cracking and local failure. Because of that, the gap elements were considered with a stiffness of
200.000 KN/m3 and they were assigned to every point of the base resulted from the meshing of the
wall. In figure 4 the interior and exterior forces which act on the rocking wall are presented.
Fig.4 - Interior and exterior forces that acts on the wall when it rocks [1]
101
In figure 4, PD1 and PD2 are the forces in the additional devices for seismic energy dissipation
that are placed at 1 and 2 from the neutral axis of the wall, PP is the post-tensioned force of the
tendons and PG is the self weight of the wall. PP and PG forces will prevent any tendency of
sliding and they will ensure the centering of the wall.
The prestressed elements will have an elastic behavior all the time during the seismic motion and
they will assure the self-centering of the wall. The initial design of the tendons was made using
equation (1) that gathers exterior and interior forces, according to [1].
(1)
In equation (1), F is the lateral force, B is the length of the wall and H is the height of the wall.
From equation (1) it resulted that there are going to be seven rows with two tendons for each
wall. The post-tensioned force for a tendon was 4638KN.
From the studies made on the rocking walls system it resulted that they do not have the capacity
to dissipate seismic energy. This is way it was necessary to add additional devices for seismic
energy dissipation. The first solution analyzed consists of milled segments. The dimensioning of
the milled segments was made according to [7].
(2)
(3)
In (2) Asd is the total area of the reinforcement disposed at the bottom of the wall in order to
dissipate seismic energy, Asp is the total area of the tendons, fpsi is the maximum strain that can
develops in the tendons, N* is the axial force at the bottom of the wall. In (3) flp is the strain of
the tendon at the limit of proportionality, dps is the distance from the last tendon to the maximum
compression area, c is the distance from the maximum compression area to the neutral axis, Lps
is the length of the tendons, u is the drift at the ultimate state and Eps is the modulus of elasticity
for the tendon. For structure C twelve rows with two milled segments (total area of 14778mm2)
were placed. Unbounded length of the milled segments is 6.20m and they were modeled with
multilinear plastic links.
The second solution used to improve the seismic behavior of the rocking walls was to add
buckling restrained braces at the bottom of the wall. The initial design of the braces was made
according to [8]. The buckling restrained braces have 6.50m length and a cross section of
60x400mm. Six rows of two braces were placed at the bottom of each wall and they were
modeled with multilinear plastic link elements. In figure 5 the model for the walls in structure C
and D is presented.
Fig.5 - Models for the walls from structures C and D
To check the real behavior of the entire structure, at both edges of each element plastic hinges
were assigned. They were deformation controlled, moment hinges and interacting axial-moment
hinges depending on the type of the element. The modeling parameters and acceptance criteria
102
for the plastic hinges were taken from [9]. The target displacement for the static nonlinear
analysis was determined using [10]. The target displacement is determined from the elastic
response spectrum based on the displacement of the equivalent single degree of freedom system.
3. Performance assessment
The plastic rotation capacity of the structural elements were determined using [10]. In figure 6
the values for the plastic rotation from the marginal beams for all structures are presented.
Fig. 6 - Plastic rotation in hinges from marginal beams
The alternative solutions have improved the behavior of the structure. Structure D led to the best
results in terms of plastic rotation. Considering the fact that the lateral force is divided equally
between the frames and the walls, the allure of the graphical representation of the plastic rotation is
constant. This is because in the area where the frames would present the maximum deformation the
walls would have the minimum ones and in the area where the walls would present de maximum
deformation the frames would compensate with minimum deformations. Structures B and C
become dual structures where the frames have an important role. For these two structures the
plastic rotations have de maximum values on the lower floors and they decrease with the increase
of the height of the structure where the walls have the minimum deformation.
The plastic rotation in the column hinges are presented in figure 7 and they were determined
using equations given [10]. In table 1 the values are shown for the maximum plastic rotation in
all four structures.
Table 1
Maximum plastic rotation in the columns hinges
umpl[rad]
ef,A[rad]
ef,B pl[rad]
ef,Cpl[rad]
ef,Dpl[rad]
0.023
0.00689
0.00713
0.00687
0.00507
Fig. 7 - Maximum plastic rotations - columns
103
The maximum values for the plastic rotation in the columns are registered in structure B where
the classical walls were replaced with rocking walls. Structures C and D led to hinges plastic
rotations less than or equal with those of the structure A.
In figure 8 are presented the drifts for the four structures which were analyzed. The alternative
solutions have improved the behavior of the structure. The lowest drifts were obtained for
solutions B and C. As can be seen in figure 8 the allure of the graphs for structures A, B and C
are specific to dual structural system. The maximum values are recorded at the bottom of the
structures. Structure D has the maximum drifts at tenth floor. For this type of structure the allure
of the representation is specific to reinforced concrete walls system where the maximum
deformation is registered at the middle height of the structure. In all four cases the allowable
value of the drift (2.5%) for the ultimate state was not exceeded.
Fig. 8 - Maximum drifts
In figure 9 a comparison between the capacity curves for the four structures is presented. The
alternative solutions, which are composed of rocking walls and additional devices for seismic
energy dissipation, led to an improvement in the global structural response. Structure D leads to
a notable increase of strength and deformability.
Fig. 9 - Capacity curves
104
4. Concluding remarks
In this paper four structures were compared in order to highlight the advantages and
disadvantages of using rocking wall system for tall buildings that are placed in seismic areas.
Pushover analysis were conducted, leading to some conclusions.
Solutions B, C and D, in which the classical walls were replaced with rocking walls, have led to
smaller plastic rotation in hinges placed in beams than those formed in structure A. Regarding the
plastic rotation formed in the columns hinges, the alternative solutions consisting of rocking walls
and additional devices for seismic energy dissipation brought an improvement by reducing them
with 30%. Solution B led to higher values for plastic rotation formed in columns hinges with 17%.
Rocking walls are capable to control the deformation pattern along the height of the structure
reducing the story drift concentrations. This advantage can prevent structural failure due to a
possible weak story mechanism.
Comparing and studying the three structures (B, C and D), the solution consisting of rocking
walls and buckling restrained braces shows the best behavior in terms of strength and
deformability of the global structural system.
References
[1]. Nor Hayati binti Abdul Hamid, Lionel Linayage - Biaxial behaviour of thin concrete walls with rocking base
connections, Journal - The Institution of Engineers, Malaysia (Volum 68, Nr.4, Decembrie 2007).
[2]. European Committee for Standardization (2002). Eurocode 1: Actions on structures Part 1-1: General actions
Densities, self-weight, imposed loads for buildings.
[3]. P100-2006 (2006). Romanian Seismic Design Code-Part I: Design Provision for Buildings
[4]. European Committee for Standardization (2003). Eurocode 2: Design of concrete structures Part 1-1: General
rules and rules for buildings
[5]. Radu Pascu Prestressed Concrete, Basis for calculation
[6]. Kutay Orakcal, Leonardo M. Massone, John W. Wallace - Analytical Modeling of Reinforced Concrete Walls for
Predicting Flexural and CoupledShear-Flexural Responses, Pacific earthquake engineering research center, 2006
[7]. J.I. Restrepo, J. Mander, T.J. Holden - New Generation of Structural Systems for Earthquake
Resistance,NZSEE 2001 Conference
[8]. Star seismic: Preliminary design of BRBF system; Use of equivalent force method, [Online]. Available:
http://www.starseismic.com.
[9]. Federal Emergency Management Agency (2000). FEMA 356: Prestandard and commentary for the seismic
rehabilitation of buildings
[10]. European Committee for Standardization (2004). Eurocode 8: Design of structures for earthquake resistance Part 1: General rules, seismic actions and rules for buildings
105
NUMERICAL MODELLING OF RIGID VERTICAL INCLUSIONS AS

REINFORCEMENTS FOR COMPRESSIBLE SOILS
Iulia-Victoria TALPOS (NEAGOE) Teaching assistant, Spiru Haret University, Faculty of Architecture, PhD
student, Technical University of Civil Engineering Bucharest, Faculty of Hydraulic Works
e-mail: talposiulia@yahoo.com
Abstract: The reinforcement of compressible soils by rigid vertical inclusions represents a
technique which is used increasingly in the last years at international level in order to replace the
traditional solutions. The method is used in order to homogenize and reduce the surface settlements
for foundation on weak soils of large civil engineering works (roads or railway embankments,
storage tanks, bridges, wind farms, industrial platforms etc.). The system is composed of rigid
inclusions and a mattress installed over the pile caps which is made of granular material. The role
of the load transfer platform is to ensure simultaneous charge both in the inclusion head and soft
soil, which makes the difference between this method and the piled raft one. The paper presents a
3D numerical model of the behaviour of rigid inclusions under dynamic loads which was calibrated
and validated on the basis of shacking table tests on small scale models. The tests were performed
at the University of Bristol (UK) in the framework of the European project SERIES (Seismic
Engineering Research Infrastructures for European Synergies). The experimental model has a
linear elastic isotropic behaviour according to the theoretical model. This paper shows the dynamic
behaviour for various configurations of inclusions and also presents comparisons between
numerical modelling results and experimental ones.
Keywords: dynamic load, shaking table, pile, 3D numerical model, load transfer platform
1. Introduction
The behaviour of reinforced materials by rigid vertical inclusions was the subject of some
important theoretical studies in the last years (e.g. Makris & Gazetas 1992 [1], Mylonakis &
Gazetas 1999 [2], Koo & all 2003 [3]). Actually the procedure is not a novelty considering the
fact that geotechnical works of reinforced soils by driven piles have been used since the mid-70's
in Scandinavian countries.
Fig. 1 - The principle of soil reinforced by rigid inclusions (Briancon [4])
Beyond the research in some European countries there are technical norms regarding this
technique, but it often requires an experimental approach of the work. Unfortunately most of the
studies were performed for the static behaviour of this system, while the dynamic one has been
very little studied.
106
This procedure aims to reinforce compressible soils in order to increase the bearing capacity and
to reduce significantly the differential settlements, at the ground level as well as at the structure
base, in safe conditions and with respect to economic aspects. The transfer of loads is done
through a system consisting of rigid inclusions and a granular platform, which ensures the
reduction and uniformity of the settlements.
The procedure is an interesting and competitive one and it successfully replaces the traditional
methods for the foundation of large-scale constructions on compressible soils. Currently the
methods of improving soft soils by rigid inclusions are used for various application areas, but
increasingly for foundations of bridges, wind farms, industrial platforms, roads or railway
embankments, storage tanks, etc.
Rigid inclusions are embedded into the compressible layer and allow load transfer by developing
friction stresses along the inclusion. The load transfer is obtained through a combination
consisting of rigid inclusions and a granular platform (mattress), system which has to ensure the
reduction and uniformity of settlements due to the arching effect. The presence of the granular
transfer mattress is making the difference between the rigid inclusions method and the pile
foundation, as the rigid inclusions are not linked to the superstructure.
The advantages of the method are: adaptability to various types of soils, the possibility of use for
a wide range of depths ranging between 2,5 and 30 meters. Moreover, it can be used in the case
of uniformly distributed loads and it allows the development of the work without soil disposal
and storage, which leads to benefits in terms of economic and environmental protection. This
technique of improving soft soils presents a higher productivity than the traditional solutions and
can be done with low costs in terms of enhanced safety.
2. Reinforcements techniques of soft soils by rigid inclusions
The system consists of rigid inclusions and a granular platform which has an important role as
the mechanisms of load transfer toward the inclusions is developed within it. This method of
reinforcing aims to ensure the reduction and uniformity of settlements due to the arching effect.
The materials used for inclusions generally have stiffness and strength characteristics higher than
those of the reinforced soil. An important aspect which characterises the inclusions is that each
inclusion is treated as a beam that can take over traction and compression stresses and also
bending and shear. In order to allow the transfer by developing friction stresses along them, the
inclusions are embedded into the compressible layer or are lightly anchored in it. The
effectiveness of the system can be enhanced by increasing the head size of the inclusions.
Depending on the technology, the inclusions can be precast (driven piles, metallic or made of
concrete, with full section or made of pipes) or cast in place (driven or bored piles, continuous flight
auger - CFA, STARSOL, vibrocompacted concrete columns, controlled modulus columns-CMC,
jet grouting, deep in place mixing, soil-cement columns, lime columns, lime-cement columns).
Fig. 2 - Schematic diagram for using drilled piles STARSOL
107
The granular platform makes the difference between this method of reinforcing and the
traditional methods as pile foundation, in which rigid inclusions are not linked to the
superstructure. By using the granular mattress, the load is applied simultaneously on the
inclusion head and on the soft soil, which leads to a reduction and uniformity of surface
settlements due to the arching effect.
The most commonly used methods for reinforcing compressible soils are: group or network of
micropiles, foundations on groups of piles or reinforced soils by rigid inclusions in contact with
a substratum to which loads are transmitted to, via a granular distribution platform.
Fig. 3 - Soil reinforcement for a pile bridge - Rion - Antirion, Greece (Auvinet [5])
3. Experimental model
In the European program SERIES (Earthquake Engineering Research Infrastructures for
European Synergies) led by ENTPE Lyon, France a project was developed that involved
achievement of experiments on small scale models subject to dynamic loads. The project was
conducted in partnership with the Technical University of Lodz, the Technical University of
Civil Engineering Bucharest and the University of Iceland. The research was developed under
the auspices of the 7th Research Framework Programme of the European Commission.
The dynamic behaviour of soils reinforced by long inclusions (piles) was modelled through the
use of shaking table from the University of Bristol, UK. The model is original due to the use of a
high number of reinforcements and to the linear behaviour materials that have been used.
Fig. 4 - Experimental model longitudinal cross-section and overview (Soubestre [6])
The model consists of a polyurethane foam matrix and inclusions made of metallic tubes. For a
more accurate reproduction of the theoretical model, the experimental model was developed
using analogous materials in order to comply with basic hypothesis: homogeneous linear elastic
108
behaviour both for the matrix and the inclusions and perfect adherence on their interface. The
characteristics of the polyurethane foam and of the inclusions are shown in Table 1.
Table 1
Characteristics of the materials used in the model
Material
Polyurethane foam
RG50/230
Inclusion
Density (m)
(kg/m3)
Youngs modulus (E)

(kPa)
Poisson's ratio (p )
48
40
0.11
7800
210000
0.3
The foam has a linear elastic behaviour up to 4 % - 5% axial strain, after which the behaviour
becomes non-linear. The dimensions of the matrix are 1.75 x 2.13 x 1.25 meters and the weight
is about 225 kg. The height was chosen in order to allow a fundamental frequency below 10 Hz
and the other two dimensions were set according to the base platform size. The inclusions have a
height of 1.3 m, with an outer diameter of 12.7 mm and an inner one of 6.2 mm. In order to
provide the condition of adherence on the interface, the holes drilled in the array for the
introduction of steel tubes are about 1 mm smaller than the diameter of the inclusions. The holes
are perpendicular to the inclusion axes, were achieved on the axis of symmetry of the section, are
parallel to each other and have a diameter of 1.5 mm.
In order to ensure the condition that the 6 axes of the model are aligned to those of the shaking
table, the experimental model was fixed on the shaking table through a base platform. The matrix
was glued to the base platform with a sheet of PVC.
Fig. 5 - The configurations tested in the experiment (Soubestre [6])
The configurations used in the experiments are the following: 35, 17 or 9 inclusions and
unreinforced foam. For each of these configurations 4 types of boundary conditions were
considered for the heads of the inclusions: clamped - free, clamped - translational, articulated free, articulated translational.
The experimental model was subjected to various types of excitations in order to analyse its
dynamic behaviour (white noise, harmonics, sinusoidal and impact waves). In order to determine
the Eigen frequencies of the model white noise excitations were used.
It was concluded that there is a correlation between the experimental results and the
homogeneous model predictions, especially for the fundamental mode system. The spectral
response analysis of the experimental model revealed the appearance of bending stresses due to
reinforcements.
4. Numerical model
109
For the numerical modelling only the configurations with 35 and 17 inclusions with clamped free
boundary conditions were considered and also the unreinforced foam matrix. The data was introduced
into the program so that the frequencies of the Eigen modes of vibration result from modelling.
A 3D numerical modelling using Midas GTS software was performed, which worked on the
principle that the response of a multi-degree-of-freedom system was assumed by the program to
a combination of single-degree-of-freedom systems. In order to comply with the experimental
conditions it was found that the materials used for modelling have a linear elastic behaviour and
the dynamic damping rate is 5 %. The peak value of response corresponding to the natural
frequency is used for the response spectrum analysis. The model was subjected to a white noise
excitation spectrum (0-100 Hz).
Fig. 6 - Fourier Transform for the acceleration on a white noise excitation (0 100 Hz)
The frequencies of the first mode of vibration (f1) were obtained for the mesh found out to be
stable. The mesh density was chosen so that the size of the elements does not influence the first
mode frequency.
The results were compared with those obtained for the theoretical model and for the
experimental one. The theoretical results were obtained by Soubestre [6] in his PhD thesis on the
basis of the experimental research. Soubestre used the homogenization method for modelling an
environment reinforced by rigid inclusions in the dynamic mode and after comparisons it was
resulted that there is a correlation between experimental results and the homogeneous model
predictions, especially for the fundamental mode system.
Table 2
Comparison of first vibration mode frequency for experimental tests, theoretical modelling and numerical
modelling
Model
f1 theoretical model
Configuration
Foam
17 inclusions
35 inclusions
4,51 Hz
5,59 Hz
5,88 Hz
f1 - experimental
tests
5,20 Hz
5,63 Hz
5,95 Hz
f1 numerical model
5,79 Hz
6,86 Hz
6,88 Hz
Fig. 7 The mesh model for the foam matrix
110
Fig. 8 - Numerical model for the foam matrix
Fig. 9 The mesh model for the 17 inclusions configuration
Fig. 10 Numerical model for the 17 inclusions configuration
Fig. 11 The mesh model for the 35 inclusions configuration
Fig. 12 Numerical model for the 35 inclusions configuration
111
5. Conclusions
The reinforcement of compressible soils by rigid vertical inclusions is a technique that can be
successfully applied in case of seismic loads. The study results show that using rigid vertical
inclusions in order to improve soft soils leads to a significant reduction of differential
settlements.
The paper presents a 3D numerical modelling of the experimental small-scale model under
dynamic loads, developed in the European program SERIES. The numerical results were
compared with those obtained using the shacking table and with theoretical ones, obtained by
using the homogenization method. The differences between theoretical and experimental results
are comprised between 0.7% and 13.26% and between experimental and numerical results are
comprised between 10.1% and 17.9%. Thus, it may be considered that the 3D model is accurate
enough and it can be further used for studying real cases of structures.
Acknowledgements
The experimental research leading to these results has received funding from the European
Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 227887,
SERIES.
References
[1] Makris N., Gazetas G, 1992, Dynamic pile-soil-pile interaction part ii: lateral and seismic response. Earthquake
Engineering and Structural Dynamics, vol. 21, pp 145162,.
[1] Mylonakis G. & Gazetas G., 1999, Lateral vibration and internal forces of group piles in
[2] layered soil. Journal of geotechnical and geoenvironnemental engineering, pp 1625,.
[3] Koo K., Chau K., Yang X., 2003, Lam S., Wong Y. Soil-pile-structure interaction under SH waveexcitation.
Earthquake Engineering and Structural Dynamics, vol. 32, pp 395415,.
[4] Briancon L., 2002, Renforcement des sols par inclusions rigides. Etat de l'art. IREX, Paris, pp 9.
[5] Auvinet G., February 1999 Geostatistical analysis of soil data on the site of Rion-Antirion bridge,
Comparative analysis of settlement of piers foundation on soil with and without inclusions. Complementary
report submitted to Geodynamique et Structure, France,.
[6] Soubestre, J., 2011 Homogeneisation et experimentation de milieux renforces par inclusion lineaires.
Application aux fondations profondes. .
112
MATHEMATICAL MODELLING OF SHOCK WAVES GENERATED BY

BLAST EVENTS AND THEIR EFFECT ON CONCRETE STRUCTURES
George-Bogdan NICA teaching assistant, Technical University of Civil Engineering Bucharest,
e-mail: bogdangeorgenica@yahoo.com
Abstract: Nowadays the explosion of bombs or explosive materials inside or near buildings generally
yields to severe damages to the structures. Explosive resistant standards and requirements are
constantly developing. This paper presents the system of differential equations coupled with
constitutive equations and equations of state that governs the transmission of shock waves through
continuous medium. Self-adjointness special problems and basic time and space integration cycles are
presented. A number of examples of concrete and reinforced concrete are presented and the
calculations are carried out in ANSYS-AUTODYN hydrocode. Results are compared with available
literature. Conclusions based on analysis results are drawn and design recommendations are presented.
Keywords: explicit analysis, self-adjoint, explosion, Navier-stokes, Euler
1. Introduction
Nowadays the explosion of bombs or explosive materials inside or near buildings generally
yields to severe damages to the structures. For example, in December 2010, a four-storey
building located in Bacau undergone severe damages due to a deflagration event taken place in a
room at the ground floor level. The top concrete slab and a lateral brick wall were completely
destroyed. A reinforced concrete beam and a precast element suffered extreme lateral deflection.
Figure 1 shows de damages referred above, where the failed top reinforced concrete slab can be
seen through the supporting props.
Explosive resistant standards and requirements are constantly developing as well as the methods
of prediction blast wave loads from both detonations of high explosives and deflagration of
vapor clouds. The techniques for the derivation of blast loads are categorized in literature into
three classes: empirical methods, phenomenological methods and first principle methods. The
simplest methods are the empirical methods, but these are the least accurate ones as they contain
correlations against experimental data with little or no physical relationship. Phenomenological
methods lie somewhere between empirical and first principle methods, and they represent only
the essential physical phenomena of the explosion, using empirical factors and relationships
obtained from experimental data. Empirical and phenomenological methods are suited to specific
applications where the parameters that define the explosion event do not depart significantly
from the conditions under which the support experiments were performed [1].
In the case of empirical methods for high explosives, the basis lies in the scaled-distance curves
for spherical and hemispherical bursts. These are known as free-air and surface scaled-distance
curves. These curves describe the incident and normally reflected overpressures and impulses as
a function of scaled-distance and together with some experimental coefficients provide the basic
data for design. In the empirical methods the most used is the Friedlander equation (Fig.2) which
characterizes the exponential decay of the pressure time-history of the blast wave, with its
positive and negative phase duration.
The most accurate and complex is the first principle approach. This is described as flow
modeling and the Computational Fluid Dynamics (CFD) programs are referred to as hydrocodes.
The governing equations and boundary conditions are taken into consideration, accounting for
interaction of the blast wave with geometry in the domain. This is how reflection prediction,
channeling and focusing in the blast wave propagation are calculated.
When CFD is used to model a blast load, the fluid is the medium through which the blast wave
propagates. The fluid is a continuum, either compressible or incompressible, with or without
113
considering viscosity effects. Usually the air is considered for hemispherical blasts and because
of high blast wave velocity the viscosity effects are not considered, except for some boundary
layers near solid boundaries.
Fig.1 - Blast effects on a four-storey flat building in

Bacau , 2010 [2]
Fig.2 - Friedlanders pressure-time history [2]
A typical high explosion consists of two phases: the detonation process of the high explosive (i.e.
TNT) and the later expansion process. During the detonation process, the detonation wave
TNT
advances outwards through the high explosive at high constant velocity ( v 6930m / s ). In
the expansion process the rarefaction wave propagates inwards on the gas behind the detonation
wave [3]. For modeling the expansion of high explosive a computational domain (region of
interest) is employed. The boundary conditions are defined at the edges of the computational
domain and where objects occur within it. These are conditions to which the fluid must conform.
For example the mass flow through solid surfaces of objects should be zero. Other boundary
conditions are the reflection condition (changing sign of the velocity), and the non-reflecting
imaginary boundaries i.e. the flow out condition the mass flow can exit the computational
domain and should not be reflected [1].
Coming back to the detonation process, because of the high detonation velocity, this process is
considered a steady state one (does not depend on time). The propagation of the detonation front
is governed only by the thermodynamics and hydrodynamics of HE [3]. The detonation speed is
constant in time and environmental factors such as pressure or temperature in the surroundings
of the HE do not have much influence on the propagation of the detonation shock. The physics
of the steady state detonation process is governed by the equilibrium equations obtained from the
shock tube. These are the steady state laws of conservation of mass, momentum and energy
across the detonation shock propagation. In literature these are known as the Rayleigh line and
Rankine-Hugoniot equations and are completed by the Chapman-Jouguet (C-J) hypothesis.
Fig.3 - Constant velocity piston-generated shock transition

form state 0 to state 1 [4]
114
Fig.4 - Hugoniot curve and Rayleigh line [3]
Conservation of mass
1 D u0 V0
D

1
0 D u1 V1
0 D u1
Conservation of momentum
P1 P0 0 D u0 u1 u0 P1 0 Du1
Conservation of energy
e1 e0
P1u1 P0u0 1 2
1 2
2
u1 u0 e1 e0 u1
0 D u 0 2
2
(1)
(2)
(3)
where D is the detonation (propagation) velocity, P,V , u, e are the pressure, specific volume,
velocity and specific internal energy of the reaction produced explosive gas right after the
detonation shock. The subscript 0 indicates the parameters corresponding variables for the
original high explosive ahead the shock. Equation (1) is called Rayleigh line and equation (3) is
derived from the conservation of energy and is called Rankine-Hugoniot curve. Calculating the
Hugoniot curve requires the knowledge of the equation of state of the explosive gas. The
equation of state can be expressed in terms of pressure and specific volume of the explosive gas.
Mathematically speaking the equation of state is the necessary complementary relationship
between the thermodynamic parameters temperature, density and pressure.
The Chapman-Jouguets hypothesis states that for a plane detonation wave to be propagated
steadily, the Rayleigh line must be tangent to the Hugoniot curve of the gaseous detonation
products at the C-J point [3].
For the expansion process the steady-state assumption is not valid. The conservation of mass,
momentum and energy together with the constitutive equation of the material and the equation of
state form the governing equations of the problem, to which specific boundary conditions are
applied.
Usually the conservation equations are written in the Eulerian frame (spatial frame) to which
total time derivatives are employed. These derivatives are a combination of the local derivative
and the convection derivative:
D
vi
Dt t
xi
(4)
The conservation equations of mass, momentum and energy in the Eulerian frame can be derived
from the first principle statements. The constitutive equation and the equation of state are
introduced
in
the
conservation
equations.
In
the
following
formulas
represent
the
density,
the
velocity
vector,
the
acceleration
vector,
, v, x, b, e, q, r , ij , , p, R,
the body forces, the internal energy, the heat flux, the heat supply density, the stress tensor
components, viscosity, strain, pressure, universal gas constant and temperature respectively. The
complete set of equations, known as Navier-Stokes equations, can be found in literature in a
multitude of expressions relating different thermodynamic parameters like enthalpy, pressure,
entropy, temperature, internal energy, free-energy (Helmholtz) etc. The following are based on
[5], and the willing reader is referred to a number of continuum mechanics books like [6]. [11]
For a more complete discussion:
Conservation of mass
Conservation of momentum
vi 0
t
x ij b 0
(5)
(6)
115
v q
Conservation of energy
e ij i j i r 0
2 x j xi xi
Constitutive equation
ij 2 ij ij
Equation of state
p
R
kk
ij p
3
(7)
(8)
(9)
Mathematically speaking one can write five conservation equations: one for conservation of mass,
three conservation of momentum equations for each spatial direction and one equation for the
conservation of energy. In these five equations the constitutive equation is introduced, which is
basically the mathematical modeling of the material behavior, relating the strain to the stress, to
obtain a set of five equations with seven parameters: three velocity components for each spatial
direction, and four thermodynamic parameters (i.e. density, pressure, internal energy and heat).
Having less equations than the number of unknowns the equation of state must be employed (i.e.
Boyles eq. of state with its specific heat expressions) to relate two of the thermodynamic
parameters to the other ones. A set of five partial differential equations with five unknowns is
obtained to which spatial and time integration is required. To this set of differential equations
boundary conditions and initial conditions are added.
In a continuum, a solution of the above equations across the whole domain is at least difficult if
not impossible. The pressure, velocity and density fields vary spatially throughout the domain as
well as temporally. Even for a given time step the domain must be discretized into equations that
apply to small volumes of fluid of a size in which it is reasonable to assume that the variables are
constant. It is this concept of discretization into control volumes that underlies computational
fluid dynamics, finding a solution for the governing equations for each control volume
simultaneously and ensuring that the solution is compatible between control volumes [1].
Another peculiarity of the set of partial differential equations presented above is that they are
non-self-adjoin. The self-adjointness differential operator problem is revealed in the convective
(the product between the velocity and spatial derivative of quantity ). The
terms like vi
xi
differential operator being non self-adjoint special treatment like operator splitting [11] or other
mathematical procedures need to be employed. These procedures usually produce changes in the
boundary conditions. More details about this procedure can be found in [11].
The spatial discretization into a set of simultaneous equations for individual control volumes
simplifies the governing equations. For example, in the case of conservation of mass, the mass
flow into and out the control volume at a given time step is equated to the change in density
within it [1]. Another simplification of the Navier-Stokes equations lies in the nature of the
explosion itself. The blast wave travels at very high speeds, and for the compressible gas case it
is useful to assume that the fluid is inviscid. For an ideal fluid, in which the kinematic and shear
viscosities are null, the material is incapable of developing shear stresses and no heat conduction
exists. Viscosity is important only close to the solid boundaries where a boundary layer is
usually formed. The Navier-Stokes equations in which an ideal fluid is employed are known in
literature as the Euler equations.
The hydrocodes for modeling blast events are in a continuous development. One of these codes
is ANSYS-AUTODYN which uses finite volume method for fluid dynamics and finite element
method for solid dynamics. The integration cycles for both Eulerian mesh and Lagrangian mesh
are taken from [12] and presented in fig.5 and fig.6, and can be easily observed that a Eulerian
116
mesh implies more sophisticated calculation. Other main aspects like cell size, computational
efficiency and element erosion (failure) are of a high degree of interest.
Fig.5 - Lagrange calculations done in each cycle

[12]
Fig.6 - Eulerian calculations done in each cycle [12]
Referring to the cell size limitations, the main assumption is that it is reasonable to consider that
variables like density, pressure and velocity are constant within the cell. Therefore, the steeper
the gradient, the smaller the computational cell size scale required in order to obtain a sufficient
level of numerical accuracy. This implies that a larger number of cells are required and both the
computational time and computational memory is required for the solution [1]. The neardiscontinuous nature of pressure in the blast wave requires a small discretizing length scale for
sufficient numerical accuracy in the solution and a too large cell size will lead to loss of
resolution of the blast wave and under predicting its peak value [1].
For computational efficiency some hydrocodes take advantage from the spherical or radial
symmetry. The limitation of cell size by the steepest gradients close to the explosion have led
developers to implement an unstructured adaptive gridding technique, in which the domain is
continuously re-meshed as the solution progress. A coarse mesh is used throughout the majority
of the domain where flow gradients are small, but the mesh is refined in the vicinity of the front
of the blast wave to allow resolution of the much steeper pressure and density gradients [1].
Examples of adaptive mesh refinement are shown in figure 7 and 8.
Fig. 7 - Adaptive mesh refinement [13]
Fig. 8 - Adaptive mesh refinement [14]
117
Another important aspect of hydrocode modeling is the element failure. Because the structural
elements are boundary conditions of the problem, their stiffness should be taken into account.
The assumption that the structure is infinitely rigid and does not respond to the blast load may be
valid since the timescale over which the blast wave acts is an order of magnitude less than the
timescale of the structural response. One exception to this is the failure of brittle facade elements
or cladding panels. Their failure represents a change in the boundary condition and subsequently
can allow the blast wave to propagate inside the building. The manner in which the evaluation is
made is defining a pressure-impulse diagram for the failing brittle element, which is initially
defined as a boundary condition. The element is monitored until a combination of peak pressure
and delivered impulse exceed the capacity of the element. For that combination the element is
considered to fail and the boundary condition is removed from the model.
2. Numerical investigations using ANSYS-AUTODYN
This paper presents a number of numerical investigations of blast modeling and is effects on concrete
structures. As the available literature is limited, the discussion will be made in a more qualitative
manner rather than a quantitative one. The first example refers to a 50cm thick and 3,00m height
reinforced concrete wall subjected to a 100kg TNT explosion at a standoff distance of 4,00m. This
example is presented in [1] for engineering reinforcement detailing purpose, and the values of peak
static overpressure and positive phase duration are calculated using empirical methods. The example
is investigated using Ansys-Autodyn capabilities, using Euler grid for the air and TNT and a
Lagrangian grid for the concrete. The reinforcement is neglected as the main interest parameters are
pressure and positive phase duration of the blast event. Proper boundary conditions are introduced
for lateral surfaces of the modeled space. The considered internal energy of the air is
e air 2,068 105 J / kg . The elements used have 10x10x10cm for the Eulerian grid and 10x10x5 for
the Lagrangian grid. Autodyn is adapting the integration step for each cycle, and its mean value is
4 103 ms . The analysis results are presented and compared to the ones presented in [1] in table 1.
Fig. 9 - Initial state
Fig.10 - Advancing pressure wave

Charge (TNT)
Distance
Thick
Height
Long. Re-bar
Transv. Re-bar
Diag. Re-bar
Concrete cover
Positive phase
duration
Peak
overpressure
Fig. 11 - Peak overpressure at 4,84ms
118
Parameter Values [1]

100 kg (6134 cm 3 )
4,0m
50cm
3,0m
Autodyn results
100 kg (6000 cm 3 )
40mm
4,0m
50cm
3,0m
-
4,83ms
4,84ms
20S 355 / 15
8S 355 / 15
16S 355 / 15
2082kN / m 2
20,82bar
20,93bar
Table 1 - Comparison between empirical and Autodyn

results
Another example is a 50 cm reinforced concrete column subjected to a TNT blast placed in

contact with the structure. As the available literature is poor this example provides only a
qualitative insight of the deformation and parameter variation. Figures 12 to 15 present the initial
state, a representative pressure diagram, a stress diagram across the direction of the transversal
reinforcement, and an absolute velocity diagram respectively.
Fig.12 - Initial state
Fig. 13 - Pressure diagram at 10 1 ms
Fig. 14 - Stress on cycle 401
Fig. 15 - Absolute velocity at 7.4 102 ms
Fig. 16 - Damage in concrete at 1.5 10 1 ms
Fig. 17 - Damage in concrete at 3 10 1 ms
119
The concrete damage model used is the available RHT concrete model in Autodyn. This damage
model is characterized by three surfaces: an elastic one, a yielding one and a residual one, as
shown in figure 18. The failure surface is defined as a function of pressure, Lode angle and strain
rate. In order to account for the material strain hardening behavior an elastic surface is used by
the RHT model. This is obtained by scaling failure surface in the radial direction. The scaling
factor is the ratio of the elastic strength to the failure one.
Pressurep
Fig. 18 - Stress limit surfaces in RHT model[16]
Fig. 19 - Strain hardening in RHT model [15]
Figure 19 illustrates the strain hardening in RHT model. The residual surface is used for taking
into account the strength of the completely crushed material. If confining pressure exists a
certain level of shear strength (due to friction of crushed particles) will be retained by the
material.
Post failure yield surfaces are obtained by interpolating the yielding surface and the residual one.
This is carried out using a subunit factor D describing the material damage. The red areas in
figures 16 and 17 correspond to a unit damage factor, namely to a complete crushed material.
3. Conclusions
Modeling explosion events becomes more and more available through available hydrocodes.
Modeling blast events requires knowledge on the material behavior to high deformation rate as
well as handling the numerical difficulties arising in the calculation. Numerical investigations
provide insight view of the variation of the parameters like pressure, plastic strain, stress etc.
Also a numerical investigation can reveal structural weaknesses like beam to column connection.
Practical measures for increasing structural behavior to explosion events can be taken. Such
practical measures can be placing span top reinforcement in the slabs, placing diagonal
reinforcement at both top and bottom ends of a reinforced concrete wall, and increasing tie
reinforcement for columns or beams. Proper detailing of these connections is indicated and
recommended by [1].
Acknowledgements
The author would like to thank the whole ANSYS team and its representative in Romania S.C.
INAS, for providing the Autodyn license needed in the numerical investigations.
120
References
[1] Cormie, D., Mays, G., 2009 Blast effects on buildings- second edition, Thomas Telford.
[2] Nica, G.B., 2011 - Current state of knowledge of the effects of blast waves on concrete structures, (research
report) , Stadiul actual al cunoaterii asupra efectelor exploziilor la structuri din beton, Raport de cercetare,
U.T.C.B.
[3] Liu, G. R. 2003 - Smoothed particle hydrodynamics- a meshfree particle method, World Scientific Publishing.
[4] Tan , H., 2008 - Physics of explosions part 1 course , The University of Manchester.
[5] Nica, G. B., 2013 - Mathematical modeling of transmission of blast waves through continuous medium,
(research report) , Modelarea matematica a transmiterii prin medii continue a undei de oc provenit din
explozie Raport de cercetare, U.T.C.B
[6] Bowen, R. 2007 - Introduction to continuum mechanics for engineers, Texas University
[7] Gurtin, M., 1981 - An introduction to continuum mechanics, Academic Press
[8] Spencer, A.J.M., 1988 - Continuum mechanics, Longman group
[9] Bonnet, J., Wood, R., 1997 - Nonlinear continuum mechanics for finite element analysis, Cambridge university
press
[10] Malvern, L., 1969 - Introduction in the mechanics of a continuum medium, Pretience hall,
[11] Zienkiewicz, O.C., Taylor, R.L., 2000 - The finite element method - Fluid Dynamics, Volume 3, 5th Edition,
Butterworth-Heinemann
[12] Gerber, B.I., Birnbaum, N.K.m Cowler, M.S., Quan, C.X. The most suitable techniques for non-linear
analysis, Century Dynamics Inc. USA
[13] Picture on http://aoss-research.engin.umich.edu/groups/admg/education.php
[14] Picture on http://blogs.mentor.com/travis-mikjaniec/blog/2012/05/18/hockey-puck-aerodynamics/
[15] Zhenguo, T., Yong, L. Evaluation of typical concrete material models used in hydrocodes for high dynamic
response simulations I.J.I.E-Elsevier.
[16] Riedel, W. , Borrvall, T., 2011 The RHT concrete model in LS-DYNA 8th European LS-Dyna user
conference , Strasbourg
121
MATHEMATICAL MODELING FOR THE SEDIMENTATION PROCESS

IN THE RESERVOIRS
Catalin POPESCU - Assistant, PhD, Technical University of Civil Engineering, Hydrotechnics Engineering
Department, e-mail:catalin.popescu@utcb.ro
Daniel Andrei GAFTOI Assistant, PhD, Technical University of Civil Engineering, Hydrotechnics Engineering
Department, e-mail:daniel_gaftoi@yahoo.com
Drago FRILESCU PhD Student, Technical University of Civil Engineering, email:dragos.fratilescu@yahoo.com
Abstract: Sedimentation is a complex process by its nature. The amount of sediment deposit and
deposit distribution pattern could be estimated quantitatively and qualitatively by using empirical
methods but also could be simulated by using mathematical methods. Empirical methods are still used
nowadays to assess the effect of sedimentation but due to the advancement of technology,
mathematical methods are more and more used as tools to simulate the sedimentation process in time.
Moreover, the accurate assessment of sediment deposition, the correct identification of the factors
that influence the sediment process, in order to calculate the correct life span of water storages and
quickly elaborate the mitigation measures, are parameters that should be correctly evaluated. In this
respect, the development of the real time tracking devices of the sediment deposition is nowadays a
problem that has attracted the attention of scientists and researchers from all over the world.
The paper analyses the results obtained by applying mathematical modeling tools to simulate the
sedimentation process in reservoirs for a particular location, but same approach may be used for
any other reservoir. As in most of the cases, the field measurements and records are scarce and the
modelers are forced to use or accept different hypothesis. An example is given in this paper.
Keywords: Mathematical modeling, reservoir sedimentation
Content
Reservoirs and artificial lakes have the main purpose of regulating inflows and, through their
storage capacity deliver volumes of water, when they are needed, for energy production
purposes, irrigation, water supply and so on. Due to the sedimentation processes the reservoir
capacity to regulate river runoff and all the other uses are diminished, and therefore this
phenomenon poses a real danger to hydrotechnical works. The understanding of reservoir
behaviour is essential and the sedimentation problems should be at the top of the list for all the
engineers working in water management and also in the main preoccupations of local authorities
that administer the reservoirs.
Sediment is the end product of erosion or wearing away of the land surface by the action of
water, wind, ice, and gravity. Water resource development projects are most affected by
sediment which is transported by water.
The total sediment yield at a specific location along a stream consists of sediment delivered to
the channel from overland areas and sediment eroded from the boundaries by the flowing water.
If there is a dam on a river, it accumulates the sedimentary materials in reservoir and decreases
its useful volume.[1] Thus sediment load should be estimated and considered in the dam design
procedure. The most accurate method of estimating the quantity of sediment eroded from the
channel boundary involves bed-material transport capacity computations, which require gauged
and recorded data.
These methods of studying reservoir sedimentation are drawn into question as request grows for
these types of analysis. Most of them are currently used in the design and management of reservoirs.
Reservoir sedimentation is primary concerned with aspects of trapped sediment, spatial distribution
and flow release. The volume of deposits represents the actual loss of capacity, so there should be
great interest in anticipating and predicting the sediment movement or bed load transport. There
122
are numerical formulas to predict the bed load transport, such examples are Van Rijn (1984) as well
as Meyer Peter Mueller (1948). There also exist a quite large number of methods to predict
reservoir sedimentation by empiric methods according to Sloff (1991). More than 22 empirical
methods based for sediment distribution exist, and the most popular ones are area-increment and
area-decrement methods (AIM, ARM). Reservoir trap efficiency and mathematical models are used
to predict the filling and the design of the dead volume.[2]
This paper analyses the particularities of a location and the data needed to be entered in the
mathematical model in order to study reservoir sedimentation in an applied environment.
Mathematical and empirical methods are usually reduced and applied to the study of a certain
reservoir or river basin in this case.
Usually there exist a number of recordings as a data base to be used for establishing the
mathematical model parameters. This data base is later used for establishing flow conditions,
border conditions and model parameters. These recordings usually consist of inflow, sediment
concentration, topographic survey, laboratory analyses, temperature recordings and all other
available recorded data. The data is usually provided by hydrometric and hydrologic
instrumentations located in the upstream, downstream and in the reservoir.
As some cases are, reservoirs are located in remote locations and are difficult to be equipped
with measuring and recording means. This scenario often results in insufficient data. This lack of
data can also be the case of an administrator being indifferent or the lack of funds, defective
measuring equipment or just plain lack of care from the administrative authorities that usually
have the river basin under care. This often translates in engineering terms in a disaster when it
comes to analysing and modelling of any long or short term phenomenon from that particular
basin. Thus in most cases data sets and recordings are incomplete.
As most reservoirs have been commissioned for a long time, sedimentation has become an issue.
This conjuncture explains the urgent need of studies on existing and in design reservoirs.
Because of the particularities of each river and reservoir, as the case of hydraulic engineering
often is, the solutions are usually tailor made, the modelling applies to that particular spot but
the method is generally applicable.
The Arges River Basin is located in southern Romania and is bounded on the north by the
Fagaras Mountains and south by the Danube River. It is located in a highly developed and
populated region (more than 3.3 million people in urban and rural industry, agriculture, forests
and natural resources). The Arges basin is one of the most important river basins in Romania due
to its high energy production potential and water supply (industrial, irrigation, population, this
also including Romanias capital city, Bucharest).
Also, this basin is one of the most developed river basins in Romania with many lakes and
complex uses like: power generation, flood mitigation, water supply.
The Pitesti reservoir is located in the mid-central river course, at the confluence of two major
rivers, the Doamnei river and the Arges river. The upstream scheme of the Arges river is well
organized, with 4 cascade reservoirs in the upstream one of which featuring Romanias tallest
concrete arch dam (166m) with a reservoir volume of 465 million m3.
On the other side, the Doamnei river has one more tributary (the Targului river). On these two
tributaries there is only one major reservoir, about 60 km upstream of the analyzed location.
Because of its nature and location Pitesti reservoir is subject to heavy sedimentation and, mainly
coming on the Doamnei tributary.
The main purposes of the reservoir are as follows:
-
Industrial water supply for: OMV - Petrom -Arpechim Pitesti oil refinery with
- a primary Tyrolean water intake, capacity of up to 6 m3/s
- a secondary pumping inlet of 2 m3/s
123
An irrigation intake, maximum capacity of up to 2.1 m3/s, set up in the left bank, serving
some 4500 ha. of cultivated land
Electric energy generation, installed power of 7.7 MW at a rated inflow of Qi = 90 m3/s,
total head of 10.5 m with a year average energy production of 19.6 GWhr /yr.
Flood mitigation: the lake was equipped (initially) with a volume of 2.3 million m3
mitigation volume
Recreational use, sport, as the right bank serves as recreational area for Pitesti city.
The reservoir was commissioned initially in 1971 with a designed capacity of about 4.5 million
m3. In 1997 the reservoir capacity was enlarged by raising 0.5m of the perimetral dyke and dam
level. Even though, the capacity of the reservoir has dropped dramatically since then, reaching an
alarming 1 million m3 at the last bathymetry survey in 2008.
In the first period of commission (1971) the reservoir had an extensive program of dredging
established through the maintenance program. Moreover, some the heavily sustained dredging
activities, were ceased since 2005.
Table 1
Statistic inflow rates of the two rivers
0.1
0.5
Doamnei
F Surface of the receiving

basin
(km2)
1836
1980
1400
1165
960
Arge
1322
1060
780
645
540
River
Max statistic inflow rates p%
The downstream water transition is provided by 4 radial gate equipped spillways, topped by crest
flap-gates.
Data for the model is supplied by the two hydrometric stations in the upstream, on each tributary
and on the downstream as well. Unfortunately only inflow data is available, as in most cases no
sediment data is available. Thus the only available data for the analysis is the actual effect of the
sedimentation process- the filling of the lake which was monitored by a few bathymetry surveys.
The objective of any suspended sediment sampling program is to develop a correlation between
water discharge and sediment load commonly called a suspended sediment rating curve.
This rating curve is normally a plot on logarithmic paper of water discharge Qw in m3/s, against
sediment load, Qs in tons/day.
Such methods of evaluating sediment concentrations for the inflow exist. Such methods are the
FAO and the USBR methods. In both methods, a flow-duration curve is sketched from the daily
water discharges. This curve is really a cumulative frequency plot that shows the percent of time
that specific discharges are equalled or exceeded for the record period.
Available methods for estimating sediment yield include a watershed rating procedure
developed by the Pacific Southwest Interagency Committee (PSIAC, 1968), the Universal Soil
Loss Equation (Wischmeier and Smith, 1965; 1978), and the Modified Universal Loss Equation
(Williams and Berndt, 1972). These methods have been tested under a variety of conditions with
mixed success. PSIAC method was modified (by Clark, 1978; and Johnson et al., 1979) to allow
the use of existing basin and soils inventory data. As the Universal Soil Loss Equation, and the
Modified Universal Loss Equation require more detailed data. The M-PSIAC method provides a
general guide to estimating total sediment yields based on the climatic and physical
characteristics of the watershed. [5]
124
After studying all the above methods it is clear that some elements of the methods can be used
but the lack of recorded sediment data or even basin sediment data is the reason why sediment
estimation was hard to achieve and thus, mathematical model calibration proved to be a
challenge
The trap efficiency method is the generic name of the method that was applied in the study, and
with further improvement may prove to be a real aid in similar cases. Rating curves supply an
empirical method to predict reservoir behavior. In this paper a mixed empirical algorithm, based
on local observations and data recorded for the mathematical model calibration was used. The
input data was the daily inflow during a year (2010). As sediment data was not available, the
input for the model had to be somehow deduced. The second set of data available was the
bathymetry survey.
The bathymetry data had a triple role:
-
Geometry for the curve linear guideline

Calibration of the model, between two successive recordings
Calibration for the upstream sediment concentration data overlapping on the
inflow
Actually from only two sets of data (inflow and bathymetry) two vital operations were
completed:
- Calibration of the upstream boundaries
- Calibration for the sediment data.
The two steps are in close relation as one implies the other and vice versa. Both have direct
effect on the model behavior and the result interpretation.
The total amount was first evaluated between the two most recent surveys. This data is
supposedly unaltered by dredging works as the administrators of the lake stated. As the data for
the inflow was only available for a year the total sediment was obtained by deducting the two
capacities and dividing by the number of years past between the surveys. As in all mathematical
modeling some acceptable simplification hypotheses have to be accepted, this being one of them.
The mathematical model.
The available data, for the completion of the mathematical model was:
-
The reservoir capacity curve evolution, over the years.
125
Fig. 1 - Pitesti reservoir sedimentation since commisioning [4]
Fig. 2 - Topobathymetric maps (with cross-sections), from the 2008 survey
From the available data a mathematical model, using the Delft3D model, was elaborated, as
follows:
Fig. 3 - Outlines of the reservoir imported into Delft3D
Establishing the land boundary, the completion of the 2D curve linear grid and bathymetry
values in each grid node, through specific operations (linear and non linear interpolation), from
the sample data provided by the bathymetry survey.
126
Fig. 4 - Optimized Delft 3d curve linear grid, observation points and border conditions
The degree of fine/coarse for the mathematical model was established, on a generic short wave
flood ( 2days peak) without sediment data.
The degree of coarseness for the curve-linear grid was determined, in order to obtain the best
ratio between the grid dimension and the model output, versus the model calculation time
Fig. 5 - Delft 3D Curve-linear grid combined with bathymetry
The only existing data, regarding the hydrologic and sediment boundary data is the hydrograph
inflow on the two tributaries and the reservoir outflow.
The mathematical model, once the graphical data was processed, had to be transformed into a
curve linear grid. With the coarseness of the grid resolved, bathymetry data was entered.
400.00
DoamneiRiver
200.00
0.00
01/Jan
20/Feb
11/Apr
600.00
31/May
20/Jul
08/Sep
28/Oct
17/Dec
ArgesRiverdownstream
400.00
200.00
0.00
01/Jan
20/Feb
11/Apr
31/May
20/Jul
08/Sep
28/Oct
17/Dec
Fig. 6 - Upstream inflow border conditions inserted into the mathematical model
127
As soon as the mathematical model geometry was solved, the parameters of the model had to be
entered. Given the properties of the curve linear grid, and the bathymetry, a minimum of 7
layers (in depth) had to be adopted for the model. Also given the geometry, a time step of 0.1min
had to be entered in order to satisfy the model requirements for horizontal advection.
ArgesRiverupstream
80.00
60.00
40.00
20.00
0.00
01/Jan
20/Feb
11/Apr
31/May
20/Jul
08/Sep
28/Oct
17/Dec
Fig. 7 - Downstream recordings during the modelled year
Boundary data for the inflow of the two tributaries was entered from data obtained from local
hydrometric stations. The series was constituted from data recordings of every day mean inflow
over the year 2010 for the upstream border.
The upstream inflows were introduced as border conditions, bearing in mind that the model only
works as long as every border stays wet throughout all the elements. A minimum inflow had to
be ensured because once the border has no inflow (even for a time step 0.1sec) the border
element will be deactivated from that time-step onward
Sediment data for the generic known sediment to be found in this specific location, meaning
cohesive sediment, was next entered into the model.
The most difficult part of the modeling, without proper data recording was to estimate and
calibrate the sediment inflow quantity. As in the site-hydrometric stations there is no data for
sediment charge of the inflow, several mathematical operations had to be accomplished.
The output of the mathematical model had to be overlapping with the recorded data
Fig. 8 - Delft 3D output on the downstream, considering the downstream condition (predetermined reservoir level)
matched with the actual recordings, after model calibration
From the hydrologic data recording, the bathymetry surveys of the lake, publications, previous
studies and so on. information underlying the mathematical model was extracted for calculation
preparation. Although there is a significant amount of data, recordings on the amount of
sediment transported does not exist, thus on the info provided by morpho-hydrological surveys
and other studies (flood risk assessment, operation rules...and so on) the amount of the sediment
128
transport data had to be estimated. This information on reservoir sedimentation was vital to
ensure the model accuracy.
There is no data recording regarding sediment inflow or outflow. The model input data was
extracted using the trap efficiency algorithm, with the following simplification hypothesis, for
assessing the total sediment quantity on the two tributaries:
1) Presuming the inflow recordings are even over the years, and no dredging had been done,
by dividing the total amount of sediments to the total water inflow over the recorded
period (1971- 2008)conducted to a combined inflow (on the two tributaries)
2) By assessing the quantity of sediment from the moment that dredging operations stopped
(2004-2008) and considering the data for the inflow as mean year.
The input for the sediment concentration data was assessed as a non linear function (polynomial
3rd degree) for the better synchronization of the sediment concentration with the inflow
Fig. 9 - Upstream border condition concentration seriers calibrated to match the inflow, lineary and polinomial
distribution of the Doamnei River
129
Fig. 10 - Upstream border condition concentration seriers calibrated to match the inflow, lineary and polinomial
distribution lf the Arges River
The results from the mathematical model are to be further used to establish the opportunity of
sediment evacuation with the use of the existing outlets, the installation of new outlets or other
means of hydro-washing of the reservoir.
References
[1] Abebaw Ayalew Andargie: Evaluation of Sedimentation Processes in the Case of Gilgel-Gibe Reservoir,
Ethiopia: Prediction and Control Measures, UNESCO-IHE MSc Thesis, Scientific Advisor P. Boeriu
[1] Chalachew Abebe Mulatu: Analysis of the Sedimentation Process for Reservoir on particular Environment
Using Delft3D. A Case Study in Koga Irrigation and Watershed Management Project; Ethiopia. UNESCO-IHE
MSc Thesis, Scientific Advisor P. Boeriu.
[2] Studiu hidrologic n bazinul hidrografic al rului Arge. INHGA, 2008;
[3] Studiu de colmatare al Acumularii Pitesti, Apele Romane 2011
[4] Sediment and erosiondesign guide, Southern Sandoval County ArroyoFlood Control Authority 1041
Commercial Dr. SERio Rancho, New Mexico 87124 2011
[5] The role of the proper data in modelling the sedimentation process in reservoirs - evaluation and suggestions. P.
Boeriu S.Moldovanu M.Margaritescu C. Draghescu F.Izvoranu C. Popescu C. N. Thilakasiri - Hydro 2013 Innbruck
130
INTEGRATED SOLUTIONS FOR IMPROVING THE ENERGETIC

PERFORMANCE OF BUILDINGS
Ana-Maria PASRE - Master student, Technical University Gheorghe Asachi of Iasi, Faculty of Civil
Engineering and Building Services, e-mail: psranamaria@yahoo.com
Paul ANGHEL - Master student, Technical University Gheorghe Asachi of Iasi, Faculty of Civil Engineering and
Building Services, e-mail: paul.anghel@yahoo.com
Nelu-Cristian CHERECHE* - Lecturer, PhD, Technical University Gheorghe Asachi of Iasi, Faculty of Civil
Engineering and Building Services, email: chereches@tuiasi.ro
Andrei BURLACU - Lecturer, PhD, Technical University Gheorghe Asachi of Iasi, Faculty of Civil Engineering
and Building Services, email: andreibu@yahoo.com
Abstract: In this study, we analyze two accessible and innovative solutions to improve energetic
performance of a building, thus: using ventilated faade in complement with earth-air heat exchangers
(EAHX) and heat pipes to improve heat transfer in EAHX. When closed, the outer skin of the
ventilated faade provides insulation from cold weather in winter and an effective ventilation channel
to complement the EAHX. The cases we studied are for the winter and summer seasons. For both
seasons, the system is useful because it can pre-heat or pre-cool the outdoor air used in the ventilation
system. In this way, the environmental impact of the building is decreased by reducing the heat loses.
In order to obtain the most efficient solution for EAHX, we are using GAEA software. The program
provides the optimal features (dimension, length, depth) and also the outlet temperature. This
temperature represents the inlet temperature for the ventilated faade. We are using the Ansys-Fluent
software to simulate the heat transfer, realized by convection and radiation. The results consist in the
potential of renewable energy, sustainability and costs of the implementation.
Keywords: ventilated facade, heat transfer, air flow, numerical simulation, heat pipes
1. Introduction
The energy used in the operation of the buildings is designed to use a healthy and comfortable
environment, for heating in the cold season, cooling in the warm season, for lighting and
ventilation. In a time when the depletion of natural resources is inevitable, we develop a central
aim to reduce energy consumption.
Previous work has been conducted by Popa et al. [6] on a dynamical study of the flow in a doubleskin facade for modified Rayleigh in the laminar regime. Also, different studies [13,14] showed the
importance of the use of the double skin faade to save energy inside the building but did not
imply other solutions of preheating or precooling the air before introducing it in the faade.
Therefore, in this study we propose a way of combining two accessible and low-cost solutions to
improve the energetic performance of buildings: ventilated faade [2,3] in complement with
earth-air heat exchangers (EAHX) and heat pipes [7,8] to improve heat transfer in EAHX [12].
The cases we studied are for the winter and for the summer season. For both seasons, the system
is useful because it can pre-heat or pre-cool the outdoor air used in the ventilation system,
meaning less energy consumption to maintain comfort for habitants.
Fig. 1 - Double glazed ventilated faade
Fig. 2 - Heat pipe
131
2. Solutions description
In order to pre-heat or pre-cool the exterior air before introducing it in the faade, we proposed
earth-air heat exchangers. To improve the heat transfer in EAHX we used gravitational heat
pipes, located in a battery. The EAHX has two parallel pipes specially designed for each season:
for the winter season the heat pipes are laid below the EAHX and for the summer season the heat
pipes are laid above, for a proper operation.
a)
Winter season
b) Summer season
Fig. 3 - Integrated solutions for improving the energetic performance of buildings, using ventilated faade upgraded
with earth-air heat exchanger and heat pipes
3. Numerical simulation
3.1. Temperature determination from EAHX
The principal process taking place in an earth-air heat exchanger is that ambient air flows
through a system of pipes which is buried in the ground. The air exchanges heat transfer through
the pipe walls with the surrounding earth.
In order to obtain the most efficient solution for EAHX, we are using GAEA software (Graphical
Design of Earth Heat Exchanger) [15]. The program provides the optimal features (dimension,
length, depth) and also the outlet temperature, for winter and summer seasons, in their cooler or
warmer day [5,9]. This temperature represents the inlet temperature for the double skin
ventilated faade. The program cannot evaluate the heat transfer through the heat pipes, therefore
we calculated the heat transfer surface for the heat pipes and we added that surface to the length
of the earth-air heat exchanger.
Fig. 4 - The variation of temperatures along EAHX for the cooler and warmer days in one year
Winter season coolest day: Tinlet EAHX = -14,2 [ C], Toutlet EAHX = -10,4 [ C],
Summer season warmest day: Tinlet EAHX = 32,5 [ C] , Toutlet EAHX = 28,9 [ C].
132
The soil used in our calculation is typical preluvisol (which is present in Bucharest area), and the
thermal conductivity was 0,12 cm2/s. The features for the EAHX upgraded with heat pipes are:
two parallel circuits with pipes (one circuit for each season), 33 m long (includes the surface for
the heat pipes), with a diameter of 300 mm, disposed at a 2 m depth.
Fig. 5 - The EAHX temperature and the annual heat exchange in one year
3.2. Numerical results

The proposed integrated solutions for improving the energetic performance of buildings involves a
double skin ventilated faade upgraded with earth-air heat exchanger and heat pipes. Using all these
elements led for the winter season to a decrease in heat loses through the interior window of the
faade with 8% for the case using the faade only with EAHX and with a decrease in heat loses with
11% for the case using the faade upgraded with EAHX and heat pipes. As for the summer season,
these solutions led to a decrease in heat gains through the interior window, therefore we obtained a
decrease of 5% for the case using the faade and EAHX and 6% for the case using the faade
upgraded with EAHX and heat pipes. All the decrease of heat loses and heat gains were calculated
by comparison with the basic case of double skin faade without any improvements.
Y axis- heat transfer in the faade

X axis- studied cases
Fig. 6 - The improvement of double skin ventilated faade using EAHX and heat pipes
3.3. Analysis of Results

We used the Ansys-Fluent software [10,11] to obtain the numerical simulations for the faade,
with the temperatures calculated with GAEA software. Therefore we could observe how the air
temperature will vary inside the ventilated faade.
The boundary conditions for the ventilated faade were: 0,2 m wide channel, uniform solar
radiation, which for the winter season is 138 W/m2 and for the summer season is 408 W/m2, and
convective heat exchange with ambient air for both glazing of the faade. For the basic case, the
inlet temperature for the faade was the temperature from the outside air and for the rest of the
133
cases the inlet temperature was the temperatures calculated with GAEA software. Laminar flow
was used in simulation with a uniform rectangular mesh.
a)
b)
c)
Fig. 7 - Double skin ventilated faade winter season
basic case,T air, inlet = -14,2[ C],T air, outlet =-12,18[ C], b) improved with EAHX T air, inlet = -11,2[ C],
T air, outlet =-9.48 [ C], c) improved with EAHX and HP T air, inlet = -10,4[ C], T air, outlet =-8.8[ C]
a)
b)
c)
Fig. 8 - Double skin ventilated faade summer season
a) basic case, T air, inlet = 32.5[ C],T air, outlet = 32,33[ C], b) improved with EAHX T air, inlet = 29,8[ C],
T air, outlet = 29,9 [ C], c) improved with EAHX and HP T air, inlet = 28,9[ C],T air, outlet =29,1[ C]
We can observe that in the winter season, the solutions of using earth-air heat exchanger and heat
pipes led to an increase of the air temperature inside the channel, so the heat transfer through the
interior window is reduced. Also, for the summer season, the solutions led to a decrease of air
temperature inside the channel, so the heat gains through the interior window are reduced.
a)
134
b)
c)
Fig. 9 - Temperature profiles for the winter season
a) basic case,T air, inlet = -14,2[ C], T air, outlet =-12,18[ C], b) improved with EAHX T air, inlet = -11,2[ C],
T air, outlet =-9.48 [ C], c) improved with EAHX and HP T air, inlet = -10,4[ C], T air, outlet =-8.8[ C]
a)
b)
c)
Fig. 10 - Temperature profiles for the summer season
a) basic case, T air, inlet = 32.5[ C],T air, outlet = 32,33[ C], b) improved with EAHX T air, inlet = 29,8[ C],
T air, outlet = 29,9 [ C], c) improved with EAHX and HP T air, inlet = 28,9[ C],T air, outlet =29,1[ C]
In the temperature profiles for winter season we can observe an improvement of air temperature
in the channel of the faade in the case of the ventilated faade upgraded with EAHX and heat
pipes than the basic case of double skin faade, which means that the heat loses are reduced. As
for the summer season, the temperature profiles show also an improvement of air temperature
inside the channel, which means that the heat gains are reduced.
135
4. Conclusions
The integrated solutions for improving the energetic performance of buildings led for the winter
season to a decrease in heat loses through the interior window of the faade with 8% for the case
using the faade only with EAHX and with a decrease in heat loses with 11% for the case using the
faade upgraded with EAHX and heat pipes. For the summer season, the integrated solutions led to a
decrease in heat gains through the interior window, therefore we obtained a decrease of 5% for the
case using the faade and EAHX and 6% for the case using the faade upgraded with EAHX and
heat pipes. The energy consumptions were reduced through the decrease of heat loss with more than
5 W/m2 in winter, and the heat gains were reduced with 4, 7 W/m2.
References
[1] N.C. Chereches, M.L. Chereches, Double skin ventilated facades, thermodynamics processes, Editura
Politehnium, Iasi, 2011
[2] Faist, A., La faade double peau : typologie et ralisations, Institut de Technique du Btiment, Lausanne
(Suisse), Rapport final EPFL, 1998
[3] Harris Poirazis, Double Skin Faades for Office Buildings - Literature Review, Lund University, Lund Institute
of Technology, Report EBD-R--04/3, 2004
[4] M. De Paepe, A. Janssens, Thermo-hydraulic design of earth-air heat exchangers, Energy and Buildings, 35
(2003), 389-397
[5] St. Benkert, F.D. Heidt, D. Scholer, Calculation tool for earth heat exchangers GAEA, Siegen, Proceedings
Building Simulation '97, Fifth International IBPSA Conference, Prague, Czech Republic, September 8-10, 2007
[6] C. Popa, D. Ospir, S. Fohanno, C. Chereches, Numerical simulation of dynamical aspecrs of natural convection
flow in a double-skin faade, Energy and Buildings, 50 (2012), 229-233
[7] Burlacu Heat transfer simulation for a hot air heating instalation with heat pipes, Juniorstav 2007, Brno,
Czech Republic, 2007
[8] Burlacu Researches regarding the efficiency of water to air heat exchanger with heat pipes for the mechanical
ventilation system, P.III -56 -60, Jubilee International Scientific Conference VSU2008, Sofia, Bulgaria, May
2008
[9] G.-E. Vlad, C. Ionescu, H. Necula, A. Badea, Thermoeconomic design of an earth to air heat exchanger used to
preheat ventilation air in low energy buildings, 2009
[10] Pardad Petro Danesh- Ansys Fluent Guide, USA, 2011
[11] ANSYS, Inc- Ansys Fluent User's Guide, Release 14.0, USA, November 2011
[12] A.-M. Pasare, P. Anghel, A. Burlacu, N.-C. Chereche, Improvement of energy performance of a double skin
ventilated faade, Mathematical Modelling in Civil Engineering, No. 4, Bucharest, December 2012
[13] D. Ospir, C. Popa, C. Chereches, G. Polidori, S. Fohanno, Flow visualization of natural convection in a vertical
channel with asymmetric heating, International Communications in Heat and Mass Transfer, Volume 39, Issue
4, April 2012, Pages 486-493
[14] Grigora, C. G. Popovici, Th. Mateescu, N.-C. Chereches, Energy saving using double skin glazed faade,
Environmental Engineering and Management Journal, 2011, in press (accepted 4rd July 2012)
[15] GAEA- Graphical Design of Earth Heat Exchange- Universitt Siegen, Deutchland, http://nesa1.unisiegen.de/index.htm?/softlab/gaea.htm
136
R2D NUMERICAL MODELING OF FLOW THROUGH GEOMEMBRANE

DEFECTS IN A LANDFILL LINING SYSTEM
Gheorghe PANTEL - PhD student, Technical University of Civil Engineering, Geotechnical and Foundation
Department, Bucharest, e-mail: george.pantel@yahoo.com
Abstract: Geomembranes are widely used as lining at the base of landfills. Their performance
depends on their integrity, which is easily affected by installation operations, resulting in a number
of defects of various sizes. An experimental program to determinate the flow through the
geomembrane defects is very time consuming, as the soils used beneath the geomembrane are
usually very low permeability clays. The advantages of a numerical modeling are the short period
of time in which it can be performed and the possibility to easily vary the input parameters of the
model. The paper presents the finite element modeling for various situations and comparisons
between numerical, experimental and empirical results.
Keywords: landfills, geomembrane, finite element, numerical modeling.
1. Introduction
Although the geomembranes are practically impermeable, after installation and during the
lifetime of a sealing system some damage can appear. Even if a quality control is implemented,
for both manufacturing and installation, still not all the defects can be eliminated.
Giroud and Bonaparte[1] presented in 1989 field data on the size and frequency of defects. These
data were obtained from quality control programs and post-operation analysis of various systems
of landfills. The main types of defects that were discovered were discontinuous welding, failure
of welds, defects caused by construction equipment or punctures caused by stones. The
frequency of occurrence of a weld defect is one for 10m of weld. This can be reduced to a defect
for 300m of weld if a quality control is carried out effectively.
Also, Giroud and Bonaparte [1] recommended the use of a defect for a 4000m2 area of
geomembrane. Although no data were available on the size of defects, they recommended the
use of a defect area of 1cm2 for the design of sealing system components, i.e. 3.1cm2 for
determining sealing system performance during normal operation.
Another recent study, which aimed to identify the position and size of defects in the
geomembrane was conducted by Nosko and Touze-Foltz [2] in 2000. They used an electric
detection system to assess the position and size of defects resulting from the geomembrane
installation. They used this system to over 300 landfills, respectively over 3 250 000m2. The
most frequent defects were caused by punctures by granular elements. Most defects were
between 0.5 and 10cm2, equivalent to a diameter comprised between 0.8 and 3.6cm. A total of
4 194 defects were detected, which corresponds to a defect for about 800m2. This value is
determined prior to making a quality control and before making any repairs. However, no matter
how well the quality control was done, still some defects may remain and after finishing the
construction new defects can appear. Therefore, the use of a defect for 800 m2 will result in
conservative predictions of infiltration through geomembrane.
For determining flow rate through defects, three methods are currently used. The first and most
simple method is to use empirical equations [3], [4], [5], but their results often do not reach the
real flow rates. A second method is an experiment that uses the exact materials that will be used
in the future location, but the major inconvenient of this method is the large duration. A third
option is to perform a numerical modeling, the results of which are close to the real flow rate and
the time requirement is shorter.
137
At present there is a number of computer software that can perform a numerical modeling for
water flow, but not all can model very well the transition from a material with a very low
permeability to one with relatively high permeability, as is the case of lining systems in landfills.
In 2001, Foose et al. [6] developed a numerical simulation of 2D and 3D finite element program
using MODFLOW. They studied the infiltration rate through a circular defect located in a
geomembrane over a clay or geosynthetic clay liner. They used a uniform mesh with a higher
density around the defect. Results obtained from modeling were compared to values resulted
from using empirical equations according to Giroud [7] and analytical models developed by
Rowe [8]. Results obtained from the 3D modeling were similar to those obtained with Rowe's
model. On the other hand, the results obtained using empirical equations developed by Giroud
overestimated the flow rate through the defect.
2. Experimental study on flow rate through geomembrane defects
The objective of this study was to determine the experimental flow rate through geomembrane
defects. The study was conducted on two types for soils, a high permeability one (sand) and a
low permeability soil (clay). In order to determine the permeability of the sand, a permeameter
was developed in order to allow the introduction of a geomembrane. The clay was tested in a
triaxial compression apparatus.
2.1. Geomembrane + high permeability material
The permeability of the sand used for the experiments, whose granulometric curve is shown in
Figure 1, was determined using a falling head permeameter. The density of the sand was of
1.78g/cm3.
Fig. 1 - Granulometric curve for sand
A total of 3 tests have been performed on the high permeability material and the results are
shown in Table 1. The resulting average permeability coefficient is 1.85 x 10-3 m/s
For determining the flow rate through the geomembrane defect installed above a high
permeability material a permeameter was used, consisting of an upper part which allows purging
the air collected above the geomembrane (figure 3(a)), a cylinder with a inner diameter of 9cm
and a 20cm height (figure 3(b)) and a lower part which is provided with a screen that prevents
flushing the material from the cylinder and allows purging the air of the sample (figure 3(c)).
138
Table 1:
Permeability results for the high permeability material (sand)
T1
T2
T3
Tm
hi
hf
(sec)
(sec)
(sec)
(sec)
(ml)
(cm)
(cm)
(m/s)
1.95
2.05
1.93
1.98
10
56.02
50.10
1.82E-03
4.07
4.20
4.08
4.12
20
50.10
44.18
1.89E-03
6.58
6.76
6.62
6.65
30
44.18
38.26
1.82E-03
9.49
9.59
9.76
9.61
40
38.26
32.34
1.83E-03
13.19
13.03
13.43
13.22
50
32.34
26.41
1.80E-03
17.34
17.43
17.69
17.49
60
26.41
20.49
1.91E-03
23.39
23.46
23.37
23.41
70
20.49
14.57
1.85E-03
32.10
32.41
32.82
32.44
80
14.57
8.65
1.86E-03
51.50
51.70
53.13
52.11
90
8.65
2.72
1.89E-03
1.85E-03
where: T1~T3 is the time measured for a level difference hi-hf; Tm is the average of T1~T3; q is
the reading on the supplying tank of the device; Hi and Hf are the levels corresponding to the
reading q and k is the permeability of the tested material.
(a)
(b)
(c)
Fig. 3 - The components of the permeameter used to determine the flow rate through the geomembrane defects:
(a) upper part (b) cylinder, (c) lower part.
This device was connected to the system that is shown in figure 4, including also a constant head
tank (figure 4(1)), two standpipes for determining the hydraulic head variation (figure 4(2)), a
permeameter (figure 4(3)) and the effluent collection system (figure 4(4)).
In this experimental program we used a high-density polyethylene (HDPE) geomembrane with a
thickness of 1mm. A circular defect with a diameter of 5mm was made in the geomembrane. The
hydraulic head on top of the geomembrane was of 24cm. The sand density was also of
1.78g/cm3 as the sand alone test. The flow rate measurement was plotted in the graph shown in
figure 5 against time. From the graph it can be seen that for approximately 5 hours the flow rate
through the system was constant, respectively 2.9 x 10-6 m3/s.
139
(a)
(b)
Fig. 4 - Experimental system for determining the flow rate through a geomembrane defect: (a) Overview (b) device
scheme (1 - constant head tank, 2 - standpipes, 3 - permeameter, 4 - effluent collection system).
Fig. 5 - Permeability test results on the geomembrane + sand system - flow rate variation against time.
2.2. Geomembrane + low permeability material

For the soil used in this series of tests the permeability coefficient was determined using a
triaxial compression apparatus. Figure 6 presents a scheme of the used system and a picture of
the triaxial cell. Figure 7 shows the results of this test. It can be seen that the material has a
permeability coefficient of 1 x 10-11 m/s.
The same HDPE geomembrane 1 mm thick was used, but the diameter of the defect was of 3
mm. The soil sample had a diameter of 10 cm and a thickness of 2 cm (Figure 6 (a)). Tests for
determining of the flow rate through the geomembrane defect were also performed in the triaxial
compression device. For this test a radial pressure of 100 kPa and an 80 kPa hydraulic head on
top of the geomembrane were applied. The device is equipped with a sensor which allows the
monitoring of the inflow water volume. Given this volume, the flow rate was determined. In the
first phase, the sample was allowed to reach saturation, for about 52 hours. After this period the
inflow water volume was monitored. In figure 8 a graph is presented showing the flow rate
passing through the defect of the geomembrane installed above a low permeability material.
140
It can be seen that, although there are variations in the flow rate, they are small enough that it can
be considered that the flow is constant.
(a)
(b)
(c)
(d)
Fig. 6 - Experimental system for determining flow rate through a geomembrane defect: (a)
geomembrane and low permeability soil sample; (b) triaxial pressure cell; (c) sample mounting;
(d) system scheme (1 water supply and pressure device, 2 pressure cell, 3 effluent
collection system, 4 automatic data collection system).
Fig. 7 - Triaxial permeability test results on clay material
141
Fig. 8 - Permeability test results on the geomembrane + clay system - flow variation over time.
It can be seen that although there are variations in the flow rate they are small enough that it can
be considered that the flow rate through the geomembrane defect is constant. The average value
that will be later compared with the values resulting from numerical modeling and empirical
equations is 1.38 x 10-10 m3/s.
3. Numerical modeling of infiltrations through geomembrane defects
3.1 Software presentation
The numerical modeling of infiltration through geomembrane defects was performed using the
SEEP2D" module of the GMS (Groundwater Modeling System), software developed by
Aquaveo.
This program was developed in the first phase by the "United States Army Engineer Waterways
Experiment Station," and later by Aquaveo. It was designed to model a number of problems
involving fluid seepage through soils. It is governed by the Laplace equation:

h
h
h
h
K xx
K yy
0
K xy
K yx
x
x
y y
y
x
where: h piezometric head;
K hydraulic conductivity.
It is a software using the finite element method to solve planar 2D problems. The models can
have defined, partly defined or undefined borders, it can also simulate saturated and unsaturated
conditions. The first modeling stage is to define the mesh that will delineate the area to be
studied. The second step is to define the boundary conditions of the studied area. These can be
hydraulic load in a given area or conditions imposed to the liquid if it is allowed to leave the
studied area or not. A third step is the actual running of the program that will calculate various
parameters such as hydraulic load, flow rate and pore pressure in each mesh node. After this step
the program generates a variety of graphs for parameters such as hydraulic load or pore pressure,
but may also provide the total flow passing through the area studied and the flow lines.
3.2 Validation of numerical models
The validation of the numerical model for the studied case, i.e. flow through geomembrane
lining systems with defects, was based on experimental studies reported in the literature, after
which the software and the model were validated based on my own tests described in chapter 2.
142
In the following paragraphs the numerical modeling based on case studies from literature and
from our own experiments is presented.
3.2.1. Numerical modeling of Barrosos (2006) experiment
Barroso Touze-Foltz, von Maubeuge and Pierson [6] conducted in 2006 an experimental
program to determine the flow rate passing through a geomembrane defect placed over a
geosynthetic clay liner installed over a layer of soil. The experiment was carried out in an
apparatus shown in figure 9.
Fig. 9 - Scheme for the apparatus used by Barroso et al[9].
The main features of this experiment were as follows:

-
thickness of the soil sample 4.8cm;
geomembrane thickness 2mm;
hydraulic heat over geomembrane 30cm;
sample diameter 20cm;
defect diameter 3mm;
soil permeability 8 x 10-11 m/s;
flow rate passing through geomembrane defect 1 x 10-11 m3/s.
Based on the above presented parameters and using the SEEP2D module of the GMS program,
a numerical modeling of the experiment was performed. In figure 10 the mesh of the studied
domain is presented. The modeled domain was only the soil sample. The bottom part of the
model was considered free draining; at the upper part, the thickness of the geomembrane was not
modeled, only its defect. At the bottom of the model a boundary condition of nil hydraulic head
was imposed. At the bottom of the model two boundary conditions were imposed. The hydraulic
143
head for the defect was 30cm and in the rest a no flow condition was imposed. For the mesh a
uniform distribution of nodes was used. The distance between two nodes was of 3mm, equal to
the geomembrane defect size and the elements have an irregular triangle shape. A total of 1299
nodes and 2430 elements were generated.
From the numerical modeling resulted that a 1.78 x 10-12 m3/s flow rate passes through the
geomembrane defect. In figure 11 the graphic representation of the flow lines can be seen and in
figure 12 the pressure head.
Fig. 10 - The mesh for the numerical modeling of Barrosos (2006) experiment
Fig. 11 - Flow lines Numerical modeling of Barrosos (2006) experiment
Fig. 12 - Pressure head - Numerical modeling of Barrosos (2006) experiment
3.2.2. Numerical modeling Chais (2005) experiment

Chai et al [7] have developed in 2005 an experiment on larger scale apparatus, more than one
meter in diameter. In Figure 13 a scheme of the apparatus and a photo of it are shown.
The main features of this experiment were as follows:
-
thickness of the soil sample 36.3cm;
hydraulic head above geomembrane 500cm;
sample diameter 117cm;
defect diameter 5mm;
soil permeability 8 x 10-8 m/s;
flow rate passing through the geomembrane defect 6 x 10-8 m3/s.
144
(a)
(b)
Fig. 13 - Scheme (a) and photo (b) for the apparatus used by Chai et al[10].
This experiment was also modeled using the SEEP2D modules of the GMS program. In figure
14 the mesh of the studied domain is presented. The boundary conditions were the same as
presented in paragraph 3.2.1. The only difference is that the hydraulic head for the defect was
500cm and the distance between to nodes was 5mm. A total of 19977 nodes and 39337 elements
were generated.
From the numerical modeling a flow rate of 2.31 x 10-8 m3/s resulted passing through the
geomembrane defect. In figure 15 one can see the graphic representation of the flow lines and in
figure 16 the pressure head.
Fig.14 - The mesh for the numerical modeling of Chais et al (2005) experiment
Fig. 15 - Flow lines Numerical modeling of the Chais et al (2005) experiment
145
Fig. 16 - Pressure head - Numerical modeling of the Chais et al (2005) experiment
3.2.3 Numerical modeling of the own experiments

After validating the software and model used to determine the flow rate passing through a
geomembrane defects located over various types of soils using the two case studies presented
above, numerical modeling for the own tests was performed.
3.2.3.1 Numerical modeling for geomembrane + high permeability material system
A numerical modeling of the experiment based on the experimental data (paragraph 2.1, figure
4) was performed, using the SEEP2D modules of GMS. For the modeled domain the boundary
conditions were the same as presented in paragraph 3.2.1. The only difference is that the
hydraulic head for the defect was 24cm. For the most part of the mesh the nodes distribution was
uniform and the distance between two nodes was 5mm. Around the defect the mesh was denser,
as shown in figure 17(a). A total of 1147 nodes and 2153 elements were generated.
From the numerical modeling a flow rate passing through the geomembrane defect of 1.09 x
10-6 m3/s resulted. In figure 17(b) the graphic representation of the flow lines can be seen and in
figure 17(c) the pressure head.
(a)
(b)
(c)
Fig. 17 - Numerical modeling for geomembrane + low permeability material system.

(a) Mesh; (b) Flow lines; (c) Pressure head.
146
3.2.3.2 Numerical modeling for geomembrane + low permeability material system

In this case the model had a total of 385 nodes and 670 elements, with a hydraulic head above
the defect of 8 cm. The boundary conditions were the same as presented in paragraph 3.2.1. The
distance between the nodes was 3mm. Around the defect the mesh was denser, as shown in
figure 18.
From the numerical modeling resulted a flow rate of 3.92 x 10-10 m3/s. Figure 19 shows the
flow lines and figure 20 the pressure head for this case.
Fig. 18 - Mesh for the numerical modeling for geomembrane + low permeability material system.
Fig. 19 - Numerical modeling for geomembrane + low permeability material system. Flow lines.
Fig. 20 - Numerical modeling for geomembrane + low permeability material system. Pressure head.
4. Comparative analysis of the results

In this analysis the results obtained from the four case studies are compared using various
methods: experimental (based on literature data and the own results), numerical and empirical
equations.
The empirical equations for low permeability soils are those developed by Giroud[8]:
- For excellent contact conditions:
Q 0.096a 0.1 hw0.9 k s0.74 1 0.1hw / H s
0.95
- For good contact conditions:
Q 0.21a 0.1hw0.9 k s0.74 1 0.1hw / H s
0.95
147
- For poor contact conditions:
Q 1.15a 0.1 hw0.9 k s0.74 1 0.1hw / H s
0.95
In the case of high permeability soils the empirical equation is the same as for orifices:
Q 0.6 a 2 gh
In tables 2 and 3 the obtained results are summarized: results of the literature experiments,
results of the own experiments, those derived from numerical modeling and the values obtained
using empirical equations. Moreover, the ratios between different values are presented:
experimental vs. numerical values, experimental vs. empirical equations respectively.
Table 2
Comparison of flow rate values obtained by various methods for the 2 case studies from the literature.
Method
Experimental
Numeric
Empiric
Barroso et al (2006) experiment

Ratio
Flow rate values
Exp./num.
3
(m /s)
Exp./emp.
10-11
1.78x10-12
5.6
-12
E: 5.26x10
1.9
G: 1.15x10-9
115
P: 6.3x10-9
630
Chai et al (2005) experiment

Ratio
Flow rate values
Exp./num.
3
(m /s)
Exp./emp.
6x10-8
2.31x10-8
0.4
-6
E: 1.71x10
29
G: 3.74x10-6
62
P: 2.05x10-5
342
Note: Exp = experimental; num = numeric; emp = empiric

E excellent contact conditions; G good contact conditions; P poor contact conditions
Table 3
Comparison of flow rate values obtained by various methods for my 2 experiments.
Experimental
Numeric
High permeability material

Ratio
Flow rate values
Exp./num.Ex
3
(m /s)
p./emp.
2.90x10-6
1.09x10-6
0.38
Empiric
O: 2.56E-5
Method
8.8
Low permeability material

Ratio
Flow rate values
Exp./num.
3
(m /s)
Exp./emp.
1.38x10-10
3.92x10-10
0.35
-9
E: 1.55E
11
G: 3.17E-7
2297
P: 1.74E-6
12608
Note (in addition to note Table 2): O empirical equation for high permeability material
It can be observed that there is a better correlation between numerical and experimental results
for Chais case study, compared to the Barrosos one. A possible source of error can be the
manner in which the soil permeability is determined. Due to the heterogeneity of soils the
permeability may vary from one section to another.
In case of empirical equations it is observed that in the first case (Barroso), for the excellent
contact conditions, the results are similar to numerical values, but for the second case (Chai) the
difference between the two values is of two orders of magnitude.
It is also noted that in both cases the numerical modeling of our own tests gave values very close
to those obtained from the laboratory tests. In the case of empirical equations, the results are
lower than the experimental values.
148
5. Conclusions
Although the best way to determine the infiltration rate through a geomembrane defect is to carry
out an experimental program using the same soil and materials as on site, this is very time
consuming and requires an available laboratory. Therefore, numerical modeling has the
advantage that is less time consuming and can simulate many situations and configurations. The
main parameters can be varied according to each location or stage of completion.
It was concluded based on the performed analysis that numerical results and the experimental
values are relatively close, the numerical model being thus validated.
References
[1] J.P. Giroud, R. Bonaparte(1989) Leakage through liners constructed with geomembrane Parts I and II
Geotextiles and Geomembranes, 8(1) and 8(2), p. 27 - 67 and 71 - 111.
[2] V. Nosko, N. Touze-Foltz(2000) Geomembrane liner failure: Modeling of its influence on contaminant
transfer Proc. 2nd European Conf. on Geosynthetics, Bologna, Italy, Vol 2, p. 557 - 560.
[3] J.P. Giroud(1997) Equations for Calculating the Rate of Liquid Migration Through Composite Liners Due to
Geomembrane Defects, Geosynthetics International, Vol. 4, Nos. 3-4, p. 335 - 348.
[4] J.P. Giroud, M.V. Khire, J.A. McKelvey(1997) Rate of Leachate Migration Through a Defect n a
Geomembrane Underlain by a Saturated Permeable Medium, Geosynthetics International, Special Issue on
Liquid Migration Control Using Geosynthetic Liner Systems, Vol. 4, Nos. 3 - 4, p. 323 334.
[5] N. Touze-Foltz, M. Barroso(2006) Empirical equations for calculating the rate of liquid flow through GCLgeomembrane composite liners, Geosynthetics International, vol 13, No. 2
[6] G.J. Foose, C.H. Benson, T.B. Edil(2001) - Predicting Leakage through Composite Landfill Liners J.
Geotech. and Geoenvir. Engrg., ASCE, 127(6), p. 534 - 539.
[7] J.P. Giroud(1997) Equations for Calculating the Rate of Liquid Migration Through Composite Liners Due to
Geomembrane Defects, Geosynthetics International, Vol. 4, Nos. 3-4, p. 335 - 348.
[8] R.K. Rowe(1998) - Geosynthetics and the minimization of contaminant migration through barrier systems
beneath solid waste Keynote Lecture, Proc. 6th Int. Conf. on Geosynthetics, Vol. 1, Atlanta, GA, p. 27 - 103.
[9] M. Barroso, N. Touze-Foltz, K. von Maubeuge, P. Pierson(2006) Laboratory investigation of flow rate
through composite liners consisting of a geomembrane, a GCL and a soil liner Geotextiles and
Geomembranes, 24(3), p. 139 - 155.
[10] J.C. Chai, N. Miura, S. Hayashi(2005) - Large-scale tests for leachate flow through composite liner due to
geomembrane defects Geosynthetics International 12, No. 3,p. 134 144.
[11] N. Touze-Foltz, J.P. Giroud(2003) - Empirical equations for calculating the rate of liquid flow through
composite liner due to geomembrane defects Geosynthetics International, 10(6), p. 215 - 233.
149
NUMERICAL MODELLING OF PILED RAFT FOUNDATIONS

rpd SZERZ Doctoral Student, Technical University of Civil Engineering of Bucharest, Department of
Geotechnics and Foundations, Faculty of Hydrotechnics, e-mail: arpad_szerzo@yahoo.com
Abstract: This paper presents some of the results of a parametric study performed on simple piled
raft foundations, modelled using three-dimensional finite elements. The information presented here
is part of a broader study conducted by the author as part of his PhD thesis [1] (not finalised at the
time of the submission of this paper) that also includes the analysis of raft foundations (in the same
configuration, geotechnical and load conditions as the piled rafts), with the primary objective of
comparing settlements and eventually the settlement-reducing capacity of friction piles placed in a
relatively rare configuration (inter-axis distance of 4.5 diameters or more).
The paper will briefly present the input data and analysis cases, before reviewing the main results
(basically settlements) of the study, through a number of comparative tables and graphs and also
by qualitative conclusions. The comparison with raft foundations will also be included for a more
in-depth picture.
Keywords: piled raft, numerical, parametric, comparison
1. Introduction
The concept of piled raft foundations is relatively new in the engineering literature and practice.
As far as geometry, constituent materials and technology are concerned, piled rafts are very
similar to pile foundations: a group of piles connected at the level of pile heads by a reinforced
concrete slab. The distinction refers basically to the modelling of the behaviour under compressive
loads. It is supposed that the pile foundations are purely indirect foundations, which means that they
transfer the vertical loads through the shaft and the tip of the piles to the soil, while the role of the
raft is limited to ensuring the load transfer from the structure to the piles and providing sufficient
punching capacity. In the case of piled raft foundations, on the other hand, it is considered that the
vertical loads are shared between the raft and the pile group, a hypothesis that greatly complicates
the calculations, but can lead to a considerably more economical design [2], [3].
In reality, most pile foundations behave as piled rafts, as the aforementioned load-sharing occurs.
There are two situations where there is practically no load-sharing:
-
Offshore foundations e.g. oil platforms, wind turbines: no contact pressure at raft base [2],
Closely placed piles with very little expected settlement: block behaviour of foundation.
The pile-raft interaction becomes significant when the piles are placed relatively far away from
each other (more than 4-5 D between the pile centres, where D is the pile diameter). The system
is most efficient when the soil could provide sufficient bearing capacity in case of a raft
foundation, but expected settlements (absolute or differential) exceed the limits.
Among the design methods, the 3D FEM can model most realistically the piled raft foundation
behaviour. All calculations presented in this paper have been performed in the Plaxis 3D
geotechnical software, using this method. The aim of the calculations is to analyse the
settlement-reducing capacity of the piles and to identify the pile-raft interaction from the
settlement diagrams.
2. Geometry, geotechnical conditions, analysis cases
The paper will focus on 12 models (named analysis cases), selected from the broader study [1],
consisting of simple raft and piled raft foundations, with variable soil conditions and pile
number/configuration.
150
2.1. Foundation geometry

Four different foundation systems were used, the first being a raft (mat) foundation, as presented
below case 1 and the other three being piled raft foundations with 16, 25 and 21 piles being
placed under the same raft used in case 1.
Fig. 1a Foundation geometry
Fig. 1b Plain view of foundations type 1-4
2.2. Soil lithology

The following three geotechnical profiles were considered for the calculations. The first profile
(named soil A) features a single layer with predefined (Plaxis 3D fine sand) and was selected
for the ease of comparison with other calculation methods in [1]. In case of soil B, a stratified
lithology was adapted, in order to highlight the effect of variable soil conditions on foundation
behaviour. For improved practicality, the typical Bucharest lithology was chosen. Soil C is a
particularly weak (very compressible) soil and was chosen to analyse the behaviour of the
foundations in the vicinity of their bearing capacity.
151
The selected constitutive law for the soil is Mohr-Coulomb. No underground water was
considered.
Table 1.
Geotechnical characteristics of soils
Nr. of
stratum
1
Nr. of
stratum
1
2
3
4
5
6
Nr. of
stratum
1
Name
Fine sand
A. Fine sand (as predefined in Plaxis 3D)

Until depth
s
[m ]
[kN/m3] [kN/m3]
[ ]
40
17
20
28
c
[kPa]
0
k0
[-]
0.531
B. Alternation of clays and sands / gravels (Bucharest characteristic lithology)

Until depth
c
k0
s
Name
[m ]
[kN/m3] [kN/m3]
[ ]
[kPa]
[-]
Fill
1
15
16
10
0
0.826
Clay
10
18
20
16
35
0.724
Course sand
15
20
21
36
0
0.412
and gravel
Clay
21
18
20
18
40
0.691
Fine-medium
25
20
21
30
0
0.5
sand
Clay
40
18
20
18
40
0.691
Name
Fill
Loess
C. Loess (Clrai typical lithology)

Until depth
s
3
3
[m ]
[kN/m ] [kN/m ]
[ ]
5
15
15
5
5
15
15
12
40
18
18
15
c
[kPa]
0
5
8
k0
[-]
0.913
0.792
0.741
[kPa]
[-]
11000
0.3
Table 1 (cont.)
E
[kPa]
3000
8000
[-]
0.15
0.4
25000
0.15
12000
0.35
20000
0.2
12000
0.4
E
[kPa]
2000
3000
6000
[-]
0.15
0.2
0.2
Fig. 2 Stratigraphy
Combining the 3 stratigraphies (A, B, C) and 4 foundation geometries (1-4) 12 analysis cases
resulted. Only vertical loads were applied, uniformly distributed at the raft level, with a value of
350 kPa (soils A and B) and 150 kPa (soil C). No underground water was considered.
152
Table 2.
Analysis
case
1A
1B
1C
2A
2B
2C
Soil
Nr. of piles
A
B
C
A
B
C
16
25
21
16
25
21
Analysis cases
Analysis
Load (kPa)
case
350
3A
350
3B
350
3C
350
4A
350
4B
350
4C
Soil
Nr. of piles
Load (kPa)
A
B
C
A
B
C
16
25
21
16
25
21
350
350
350
150
150
150
2.3. Other input conditions:

-
All elements of the foundation system (piles, raft and soil) were modelled as threedimensional finite elements (15 nodes), with a medium-dense mesh within the perimeter of
the raft and a course mesh outside the raft; the building itself was modelled as an equivalent
linear-elastic material.
The size of the model was chosen as 100 m x 100 m x 40 m, where 40 m is the depth of the
model.
The connection between the piles and the raft was considered perfectly rigid.
3. Results of analysis
3.1. Maximum settlements
All 12 models were analysed using the Finite Elements Method in Plaxis 3D (geotechnical
software). Following the analysis, the following values were obtained for settlements:
Table 3.
Maximum settlements of pile / piled raft foundation
Maximum settlement [mm]
Analysis
case
A
B
C
254
288
1100
1
101
159
415
2
105
138
320
3
110
150
353
4
Fig. 3 - Maximum settlements of foundations
153
Observing fig. 3, one can be note that the settlements are generally consistent with the
expectations, as follows:
-
Settlements indirectly proportional to the pile number and modulus of elasticity of the
soil layers around and below piles,
Significant reduction of the settlements when using piles.
First of all, in case of piled rafts (cases 2-4), the settlements are indirectly proportional to the
number of piles, reaching their maximum value in the 16 pile configuration and the minimum in
the 25 pile configuration. The only inconsistency occurs in case of soil A, which can be a
consequence of the relatively unrefined (medium-dense) mesh. A finer mesh that would have
refined the results would lead to a running time several times longer.
3.2. Settlement spread
Another important aspect is the spread of settlements for the same soil, for different pile
configurations. One can observe on fig. 3 that the largest difference occurs in case of soil C,
while the smallest (almost negligible) difference occurs at soil A. While it may not be a
general rule, the results suggest that the spread of settlements is directly proportional to the
settlements themselves, i.e. the settlement reduction obtained when additional piles are installed
is more pronounced in case of larger settlements (which are obtained as a combination of
difficult soils and/or large loads).
Table 4.
Settlement reduction (piled raft vs. raft foundation)
Analysis
case
2
3
4
Soil A
60%
59%
57%
Settlement reduction (%)

Soil B
45%
52%
48%
Soil C
62%
71%
68%
Fig. 4 - Settlement reduction (%)
Fig. 4 presents the extent of settlement reduction when using piles in addition to the raft
foundation. The values range from 45% to 71%, which is a very significant reduction in
settlements 11 out of 12 cases show a settlement reduction to less than a half.
Also, one can note that the largest reduction occurs in case of soil C, where the settlement of the
raft foundation was very large (more than 1 m) and the foundation was close to exhausting its
bearing capacity, with settlements already far exceeding the values normally allowed for
serviceability state.
154
3.3. Pile-raft interaction

Although a rigorous analysis of this aspect is a very complex task and involves the interpretation
of contact pressures (between raft and underlying soil), skin friction and tip resistance (between
piles and soil) and other indicators, a qualitative analysis can be carried out by interpreting the
settlement diagrams obtained from the models. The following graphs will present these diagrams
for cases 2A, 3A and 4A.
Diagrams for case 2A
a
b
c
Fig. 6 Settlement diagrams for case 2A; 6a vertical section through foundation centre, 6b settlements at pile
base (whole model), 6c settlements at pile base (close-up)
The analysis in case 2A featured 16 piles placed in a uniform grid of 4x4 with an inter-axis
distance of 6 m (6 D). Literature [2] suggests that when friction piles are placed at a distance of
(5-6) diameters, the expected behaviour is of a piled raft foundation, where the piles and the raft
settle differently and the group effect of the piles is virtually absent.
Indeed, one can observe on fig. 6c that the piles and the surrounding area have a significantly
larger settlement than parts of the raft between piles (100 mm compared to around 70-80 mm),
155
which implies higher contact pressures between the raft and the underlying soil and a higher
share of the loads by the raft.
a
Fig. 7 Settlement diagrams for case 3A; 7a vertical section through foundation centre, 7b settlements at pile
base (whole model), 7c settlements at pile base (close-up)
In this case the raft was reinforced by 25 piles placed in a uniform grid of 5x5 with an inter-axis
distance of 4.5 m (4.5 D), which is traditionally considered the value below which the group
effect of piles becomes significant. Accordingly, the influence areas of the piles intersect and the
difference between the maximum and the minimum settlement is lower than in case 2A.
156
8a
Fig. 8 Settlement diagrams for case 4A; 8a vertical section through 5 piles, 8b vertical section through 5 piles
(along the diagonal), 8c vertical section through 3 piles (foundation centre), 8d settlements at pile base (whole
model), 8e settlements at pile base (close-up)
Case 4A is a hybrid between block behaviour (dense pile configuration) and independent
behaviour (rare pile configuration). It was obtained by adding another 5 piles around the centre
of the raft to the 16 uniformly placed piles from 2A (16 piles). This created a central area with
piles placed at 3 D and a marginal area with piles at 6 D.
157
One can observe on fig. 8c that the central area shows a block behaviour, while the marginal
piles will settle independently. According to expectations, the block behaviour in the centre is
more predominant than in case 3A, the piles having very small extra settlement compared to the
area between them. Practically, the whole central area will act as one rigid body, similarly to an
individual pile.
4. Conclusions
The results presented above confirm that three-dimensional finite element analysis is suitable for
the rigorous design of piled raft foundations, as it can realistically model the pile-raft interaction,
which basically differentiates these foundations from purely indirect pile foundations.
Figures 6-8 show that the different behaviours (block behaviour, independent behaviour of piles
and intermediate situations) can be clearly identified by the settlement diagrams.
In addition, the study provides a number of settlement values for piled rafts and the
corresponding raft foundations, which can be used in various ways, e.g. to compare the
settlement reduction in case of additional piles or validate results obtained by other design
methods.
As a general conclusion, the expected results were mostly confirmed, validating the opinions
expressed in the scientific literature about the features, strong points and weaknesses of threedimensional FEM modelling.
References
[1] Szerz, ., Research Report nr. 2, 2013 (unpublished, part of future PhD thesis) Study of the Behaviour of
Piled Raft Foundations
[2] Katzenbach, R., Arslan, U., Moormann, C., Piled Raft Foundation Projects in Germany, J.A. Hemsley,
Design Applications of Raft Foundations, Thomas Telford Publishing, 2000, p. 323-392
[3] Poulos, H.G., Practical Design Procedures for Piled Raft Foundations, J.A. Hemsley, Design Applications
of Raft Foundations, Thomas Telford Publishing, 2000, p. 425-468
158
INFLUENCE ASPECTS OF EXTERNAL PARAMETERS UPON

EFFICIENCY OF A NATURAL SMOKE EXHAUST SYSTEM
Andrei-Mihai STOICA - PhD, Technical University of Civil Engineering, Faculty of Building Services,
e-mail: andrei_stoica1984@yahoo.com
Abstract: With the aim of evaluating the capabilities of a supermarket exhaust system to control
the spread of smoke in the fire scenario, thereby providing conditions for safe evacuation of people
from the building and the assurance of the necessary conditions for a safe intervention, an influence
of the external parameters such as temperature is necessary. The Fire Dynamics Simulator program
is specialized for large eddy simulations and is used to investigate the smoke filling process in a
large building. Initially, the model is used to simulate the smoke influence on the fire detection
system of the building under different fire scenarios. The performance of the smoke exhaust system
in a real hypermarket is then studied. Smoke exhaust processes are investigated under different
atmospheric conditions and heat release rates. Simulated results show the link between the external
atmospheric conditions and the evacuation flow of the smoke. Results indicate that high ceiling
temperature slows down the so-called smoke ceiling jet moving horizontally at the ceiling towards
the walls, therefore the smoke particles are remaining longer in the upper part of the building, while
low ceiling temperature accelerates such smoke ceiling jets, but in the same time cools down the
smoke particles.
Keywords: temperature, smoke hatches, fire plume, CFD, smoke flow
1. Introduction
In recent years, the hypermarket buildings, although not a new architectural design feature, have
become very popular. Many have been built in all the cities around the globe, and the concept of
large space buildings has been widely adopted for these types of commercial constructions.
Beside the technical problems associated with lighting, air conditioning and ventilation
engineering, fire protection has to be considered carefully. Note that due to the volume of a
hypermarket, the fire design strategies and regulations used for the traditional buildings, seem
inappropriate. There are though some existing fire safety design guides and codes for large space
buildings [1]-[5].
For the same reasoning, not much research work and experimental investigations on full-scale
buildings and their smoke exhaust system have been carried out. Few large-scale experiments
were reported, such as atrium experiments on smoke extraction [6]-[9], in-situ measurements on
actual large space buildings [10]-[11], and smoke hazard in large spaces [12]-[14]. The results of
these experimental studies provide raw data for understanding the smoke filling process inside a
large space under real fire scenarios.
The numerical modeling of the fire environment plays an important role in investigating the
smoke filling process in such buildings. The zone models were first used due to their relative
simplicity, allowing the observation of different phenomena in specific zones [15]-[16]. The
code used for the zone modeling is developed for buildings of specific shape and styles. If these
aspects are not according to the standard style, then the code-type equations are not suitable. The
CFD models such as Fire Dynamic Simulator do not have such limitations and are capable of
describing accurately the complex geometry of a building with many small volume elements.
The models solve the fundamental equations of fluid dynamics and heat transfer, and provide
detailed information on the fluid motions. While the use of these models has been limited by the
requirement of large computing power, modern computers have made them more accessible to
fire safety designers. Sinclair [17] successfully adopted a CFD model to simulate fires in atria of
different shapes and presented the descending features of the smoke. Ding et al. [18] investigated
the system for both natural ventilation and smoke control through experiments and CFD
159
predictions, while Chow et al. [19] studied the fire spread characteristics and smoke movement
characteristics in large space buildings experimentally and numerically [20].
When a fire breaks out on the floor of a large space building, such as a hypermarket, the hot
smoke plume rises to the ceiling by buoyancy force. While the smoke plume rises, relatively cool
ambient air is entrained into the plume to reduce the temperature, but increases the mass flow
rate of the contaminated air. When the smoke plume reaches the ceiling, the momentum of the
plume is diverted into a jet, spreading the smoke over the ceiling area. Once the ceiling jet covers
the whole ceiling, the depth of the smoke layer increases until either the entire building is filled
with smoke or the rate of smoke that entries into the layer is balanced by the rate of the exhaust
smoke. Therefore, the capacity of the exhaust system to retain the smoke layer above the
emergency escape routes is an important issue for the fire safety designers. In most large space
buildings, smoke is exhausted via the top of the roof, by natural ventilation [21].
For a large space building such as a hypermarket, fire safety design is particularly important as it
affects a large number of human lives, especially during the holidays. The smoke management
system has to ensure that the bottom of the smoke layer is kept as high as possible and for as
long as possible.
So far, most investigations have focused on the fire spread and smoke movement in atria and very
few have been done the influence of external parameters upon the efficiency of a natural smoke
exhaust system in a hypermarket. This paper focuses on the performance of natural smoke exhaust
method in different atmospheric conditions. The smoke filling process is investigated, using the
Fire Dynamics Simulator (FDS) code, which is based on Large Eddy Simulation (LES).
2. Mathematical model
The mathematical modeling of a smoke exhaust system is made possible only by solving a system
of governing equations of fluid mechanics and heat transfer. The fundamental equations of fluid
dynamics results from applying the following universal laws of conservation to a fluid flow:
- Conservation of Mass
- Conservation of Momentum
- Conservation of Energy
In addition to the equations that results from these universal laws, it is also necessary to establish
relationships between fluid properties in order to close the system of equations, such the equation
of state and chemically reacting flow.
2.1. Continuity Equation
The equation of continuity represents the form in which the Conservation of Mass law from the
general mechanics is described in fluid mechanics. The equation can more easily be obtained by
starting of the one-dimensional flow (the current tube), and then applied to a tridimensional flow.
Thus, the mass ratio that results is
(1)
and the equation of continuity in the most general form
(2)
160
Applying the Conservation of Mass law to a fluid with a density variable in time and space,
and the interior mass in the control volume
which varies in the time t, then the equation
takes the form:
(3)
where is the fluid density and V is the fluid velocity [22]. The first term of the equation
represents the rate of increase of the density in the control volume, while the second term
represents the rate of mass flux passing out of the control surface (which surrounds the control
volume) per unit volume.
It is better to use the substantial (total) derivative
(4)
To change Eq. (3) into the form
(5)
Eq. (3) was derived using the Eulerian approach. In the Euler system are followed the particles
that pass through the same point in space; more precisely, the velocities of all the particles that
pass through a specific point in space are measured, and in this conditions, the velocity becomes
a function of the point coordinates x, y, z and of time t: V=f(x,y,z,t); thus, the acceleration, as a
derivative of velocity by time, has a more complex expression [23]:
(6)
In the alternative Lagrangian approach, the changes to the properties of a fluid element are
recorded by an outsider moving with the particles [22]. The Eulerian approach is used more
commonly in fluid mechanics. For a Cartesian system, where u, v, w are the x, y, z components
of the velocity, Eq. (3) becomes
(7)
2.2. Momentum Equation

Newtons Second Law applied to fluid passing through an infinitesimal, fixed control volume
gives the following momentum equation [23]:
(8)
The first term in this equation represents the rate of increase of momentum per unit volume. The
second term represents the rate of momentum lost by convection through the control surface. Note
that
is a tensor, so that
is not a simple divergence. This term can be also written as:
(9)
When this expression for

is replaced in Eq. (8), and the resulting equation is simplified
using the continuity equation, the momentum equation reduces to
(10)
161
The first term in the right-hand side of Eq. (10) is the body forces that act from a certain distance
and apply to the entire mass of the fluid such as the gravitational force. In this case
.
The second term in the right-hand side of Eq. (10) represents the surface forces. These forces are
applied by external stresses that consist in normal and shearing stresses and are represented by
the components of the stress tensor .
(11)
where
represents the viscous stress tensor [24].
Upon substituting Eq. (11) into Eq. (10), the famous Navier-Stokes equation is obtained:
(12)
For a Cartesian coordinate system, Eq. (12) can be separated into the following three scalar
Navier-Stokes [23]:
(13)
2.3. Energy Equation

The First Law of Thermodynamics applied to a fluid passing through an infinitesimal, fixed
control volume yields the energy equation [23]:
(14)
where is the total energy per unit volume given by

energy per unit mass.
and e is the internal
For a Cartesian coordinate system, Eq. (14) becomes
(15)
which is in conservation-law form. Using the continuity equation, the left-hand side of Eq. (14)
can be replaced by the following expression:
(16)
162
which is equivalent to
(17)
if only internal energy and kinetic energy are considered significant. Forming the scalar dot
product of Eq. (10) with the velocity vector V allows obtaining
(18)
Now if Eq. (16), (17) and (18) are combined and substituted into Eq. (14), a useful variation of
the original energy equation is obtained:
(19)
The last two terms of this equation can be combined into a single term, since
(20)
This term is also called the dissipation function and represents the rate at which mechanical
energy is expanded in the process of deformation of the fluid due to viscosity. After inserting the
dissipation function, Eq. (19) becomes
(21)
Using the definition of enthalpy,

and the continuity equation, Eq. (21) can be rewritten as
(22)
2.4. Equation of State

In order to close the system of fluid dynamic equations it is necessary to establish relations
between the thermodynamic variables (p, , T, e, h) as well as to relate the transport proprieties
(, k) to the thermodynamic proprieties.
For most problems in gas dynamics, it is possible to assume a perfect gas. A perfect gas is defined
as a gas whose intermolecular forces are negligible, and obey the perfect gas equation of state
(23)
where R is the gas constant [2]. For other problems it is possible to assume a calorically perfect
gas. A calorically perfect gas is defined as a perfect gas with constant specific heats. In a
calorically perfect gas, the specific heat at constant volume cv, the specific heat at constant
pressure cp, and the ratio of specific heats all remain constant, and the following relations exist:
For air at standard conditions, R=287 m2 /(s2 K) and =1.4.

The coefficients of viscosity and thermal conductivity can be related to the thermodynamics
variables using kinetic theory. For examples, Shuterlands formulas for viscosity and thermal
conductivity are given by
163
where C1-C4 are constants for a given gas. For air at moderate temperatures, C1 = 1.458 X 10-6
kg/(msK), C2 = 110.4 K, C3 = 2.495 X 10-3 (kg m)/(s3 K3/2) and C4 = 194 K. The Prandtl
number
is often used to determine the coefficient of thermal conductivity k once is known. This is
possible because the ratio (cp/Pr), which appears in the expression
is approximately constant for most gases. For air at standard conditions, Pr = 0.72.
2.5. Chemically reacting flows
The assumption of a calorically perfect gas is valid if the intermolecular forces are negligible and
the temperature is relatively low. As the temperature of the gas increases to higher values, the gas
can no longer be considered calorically perfect. At the beginning, the energy resulted from the
vibration of the molecules and the specific heats cv and cp are no longer constant but function of
temperature. For air, this occurs at over 800K, when it becomes at first thermally perfect, which
means that its specific heats are functions only of temperature. If the temperature continues to
increase, chemical reactions start to take place, and the gas is no longer thermally perfect.
For most chemically reacting gases, it is possible to assume that the intermolecular forces are
negligible, and each individual species obeys the perfect gas equation of state. The equation of a
mixture of perfect gases can be written as
(24)
where R is the universal gas constant [8314.34 J/(kg mol K)] and M is the molecular mass of the
mixture of gases. The molecular mass of the mixture can be calculated using
M
where ci is the mass fraction of species i, and Mi is the molecular weight of each species.
The species mass fractions in a reacting mixture of gases are determined by solving the species
continuity equations, which are given by
(25)
where i is the density of each species, Ui is the diffusion velocity and

is the rate of
production of species i due to chemical reactions. The species mass fraction is related to the
species density by
If i is replaced with ci and the global continuity equation (3) is employed, the species
continuity equation can be rewritten as
(26)
The mass flux of species i(iU) due to diffusion can be approximated for most applications using
Flicks law:
where Dim is the multi component diffusion coefficient for each species.
164
The rate of production of each species

is evaluated by using an appropriate chemistry model
to simulate the reacting mixture. A chemistry model consists of m reactions, n species and nt
reactants and can be symbolically represented as
where l,i and l,i are the stoichiometric coefficients and Ai is the chemical symbol of the ith
reactant. For example, a widely used chemistry model for air is due to Blottner et al. [25] and is
composed of molecular oxygen (O2), atomic oxygen (O), molecular nitrogen (N2), nitric oxide
(NO), nitric oxide ion (NO), atomic nitrogen (N) and electrons (e), which are reacting
according to the chemical reactions
O
N
N
NO
NO
N
N
M
M
N
M
O
O
O
2O
2N
2N
N O
O
NO
NO
M
M
N
M
N
N
e
where M1, M2 and M3 are catalytic third bodies. This chemistry model involves 7 reactions (m
= 7), 6 species (n = 7) excluding electrons, and 10 reactants (nt = 10).
Once the chemistry model is specified, the rate of production of species i can be computed using
the Law of Mass Action [26]:
(27)
where Kfl and Kbl are the forward and backward reaction rates for the lth reaction and j is the
mole-mass ratios of the reactants, defined by
where Zj-n,i are the third-body efficiencies. The forward and backward reaction rates are
functions of temperatures and can be expressed in the Arrhenius form as
For the present air chemistry model, the constants for each reaction (K1, K2, K3, K4, K5, K6)
and the third-body efficiencies (Zj-n,i) are given by Blottnet et al. [25].
3. Results and discussions
The building chosen to analyze the effects of the atmospheric conditions and the influence of the
heat release rate upon the natural exhaust system is hypermarket with the dimensions 60 m long,
50 m wide and 8 m height with a total volume of 24000 m3. The natural exhaust system is
composed of 16 smoke hatches placed in the roof of the building (Fig. 15). The compensation air
enters through the exit doors.
165
Fig. 15 Description of the building
The paper tries to determine the influence of the external parameters such as the temperature of
the air and the internal conditions, such as the heat release rates and the fire position in the
building, upon the smoke detection system and the exhaust system of the building. The 32 cases
that were investigated consisted in the variation of:
a) External temperature: te = -15, 0, 15 and 30 oC
b) Heat release rate: Q = 500 W/m2 and Q=1500 W/m2
c) Position of the fire in four different places of the building
(a)
(b)
(c)
(d)
Fig. 16 The fire location
Fires of 500 and 1500 kW were simulated to determine the impact of the heat release rates on the
smoke movement. The fire was placed in four different areas (Fig. 16).
The initial temperature inside was considered 20oC. When the fire starts, the smoke plume rises
towards the ceiling because of the buoyancy. Once it reaches the roof, it forms a ceiling jet and
spreads in all the directions, covering the roof. During the fire, the compensation air enters the
buildings with speeds of 1-3 m/s. The cold air enters the smoke plume which increases its
volume, but also decreases the temperature of the gases. Fig. 17 shows the distribution of the
speed in the building after 300 s since the fire ignited.
166
Fig. 17 Distribution of the speed (m/s) after 300 s
In Fig. 18 it is easy to see the influence of the heat release rate upon the time needed for the detection
system to react.
(a)
(b)
(c)
(d)
Fig. 18 The variation of the time needed to open the smoke hatches for different external temperatures and heat
release rates
The position of the fire does not influence as much the time for the hatches to open, except the
cases 13 to 16, when the fire is placed near the cold air inlets (Fig. 19). One explanation is that
the fresh air has a bigger influence on the smaller heat release rate.
(a)
(b)
Fig. 19 Time needed to open the smoke hatches
167
Also, for a variation of the external temperature of T = 45oC, the difference of the time to open the
hatches is not significant (t 10 s) in the case of the higher heat release rate. As you can see in Fig.
20, the heat release rate does not influence the smoke flow through the hatches regardless of the external
temperatures.
(a)
(b)
(c)
(d)
Fig. 20 Smoke flow for different external temperatures and heat release rates
4. Conclusions
The FDS code is used to simulate the smoke behavior in a hypermarket building. Thirty-two
cases are investigated for different values of the internal and external parameters. Comparing
with experimental values from the literature, the values for Cs = 0.2 and Prt = 0.5 were chosen
for the numeric modeling of such large space building. The simulations determined the influence
of the external temperature on the natural exhaust system in different fire scenarios.
For external temperatures higher than 15oC, as in the case of spring and summer, the efficiency
of the exhaust system is dependent mostly on the initial sizing, and the variation of the internal
parameters does not affect too much the smoke flow out of the building.
References
[1] National Fire Protection Association 92B, Guide for Smoke Management Systems in Malls, Atria and Large
Area. Quincy, Mass., USA, 1998.
[2] Hansell, G. O., Morgan, H. P., and Marshall, N. R., Smoke Flow Experiments in a Model Atrium, BRE
Occasional Paper OP 55, Fire Research Station, Building Research Establishment, U.K., 1993.
[3] Building Code of Australia, Australian Uniform Building Regulations Co-Ordinating Council, Australia, 1990.
[4] AS43911999, Smoke Management SystemsHot Smoke Test, Standard Australia, 1, The Crescent,
Homebush, NSW, Australia, 1999.
[5] Klote, J., and Milke, J., Design of Smoke Management Systems, ASHRAE Publication 90022, 1992.
[6] Yamana, T., and Tanaka, T., Smoke Control in Large SpacesPart 2, Fire Science and Technology, vol. 5, no.
2, pp. 4154, 1985
168
[7] Huo R, Li YZ, Jin XH, Fan WC. Studies of smoke filling process in large space building. Journal of Natural
Disasters, 9 (1), pp.8992, 2000
[8] Huo R, Li YZ, Jin XH, Fan WC. Studies of smoke filling process in large spaces. Journal of Combustion
Science and Technology, 7(3):2202, 2001.
[9] Chow WK, Li YZ, Cui E, Huo R. Natural smoke filling in atrium with liquid pool fires up to 1.6MW. Building
and Environment, 36(1):1217, 2001.
[10] Dillon, M. E., Case Study of Smoke Control System Testing for a Large Enclosed Stadium, ASHRAE
Transactions: Symposia, vol. 100, no. 2, pp. 878892, 1994.
[11] Chow,W. K., Atrium Smoke Exhaust and Technical Issues on Hot Smoke Tests, Paper HT200572412, ASME
2005 Summer Heat Transfer Conference, July 1722, 2005, San Francisco, Calif., USA, 4 pages, 2005.
[12] Lougheed, G. D., Hadjisophocleous, G. V., McCartney, C., and Taber, B. C., Large-Scale Physical Model
Studies for an Atrium Smoke Exhaust System, ASHRAE Transactions, vol. 105, pp. 676698, 1999.
[13] Lougheed, G. D., and Hadjisophocleous, G.V., The Smoke Hazard from a Fire in High Spaces, ASHRAE
Transactions, vol. 107, pp. 720739, 2001.
[14] Hadjisophocleous, G. V., Fu, Z., and Lougheed, G. D., Experimental Study and Zone Modeling of
SmokeMovement in a Model Atrium / Discussion, ASHRAE Transactions, vol. 108, pp. 865871, 2002.
[15] Chow WK, Cui E. Plume equation for studying smoke-filling process in atria with a zone model. Fire and
Materials, 21:23544, 1997.
[16] Yi L, Chow WK, Li YZ, Huo R. A simple two-layer zone model on mechanical exhaust in an atrium. Building
and Environment, 40(7):8691010, 2005.
[17] Sinclair R. CFD simulation in atrium smoke management system design. ASHRAE Transactions 2001; 107:18.
[18] Ding WT, Minegishi YK, Hasemi YJ, Yamada TY. Smoke control based on a solar-assisted natural ventilation
system. Building and Environment 2004;39(7):77582.
[19] Chow WK, Cui E, Li YZ, Huo R, Zhou JJ. Experimental studies on natural smoke filling in atria. Journal of
Fire Science 2000; 18(2):85103
[20] Qin T. X., Guo Y. C., Chan C. K., Lin W. Y., Numerical investigation of smoke exhaust mechanism in a
gymnasium under fire scenarios. Building and Environment, 41, pp. 1203-1204, 2006.
[21] Milke JA. Effectiveness of high capacity smoke exhaust in large spaces. Journal of Fire Protection Engineering
2003;13:11128.
[22] Tanehill JC, Anderson DA, Pletcher RH. Computational fluid mechanics and heat transfer. 2nd ed. Washington:
Taylor&Francis, 1997.
[23] Mateescu Cristea. Hidraulica. 2nd ed. Bucureti: Editura Didactic i Pedagogic, 1963.
[24] McGrattan K, Simo Hostikka, Floyd J. Fire dynamics simulator. Technical Reference guide. Washington: U.S.
Government Printing Office, 2007.
[25] Blottner F. G., Johnson M. and Ellis M. Chemically Reacting Viscous Flow Program For Multi-component Gas
Mixtures. Sandia Labs, Albuquerque, New Mexico, 1971.
[26] W. G. Vincenti and C. H. Kruger, Introduction to Physical Gas Dynamics, Krieger, Malabar, Florida, 1976.
169
DYNAMICS OF FREE SURFACE FLOW AROUND A CYLINDER

VISUALIZATIONS AND PIV MEASUREMENTS
Nicoleta Octavia TNASE Assistant univ., PhD, University Politehnica Bucharest, Faculty of Energetics,
e-mail: notanase@yahoo.com.
Florin BODE Lecturer, PhD, Technical University of Cluj Napoca, Faculty of Mechanical Engineering,
e-mail: florinbode@gmail.com.
Abstract: Two different experimental techniques are used in this article to analyze a free-surface
flow around an immersed cylinder in a channel: a flow visualization technique using a dye and a
nonintrusive optical flow measurement method, Particle Image Velocimetry (PIV). On one hand,
visualizations allow a qualitative analysis of the flow field and the evolution of the free surface
relative to the cylinder surface, on the other hand, PIV measurements allow a quantitative
evaluation of the flow field. Two types of visualizations were made: using a photography technique
with a blue colored dye and with the same 2D PIV system used for the measurements. The occurrence
and development of the instabilities of the flow, downstream the cylinder, and the zone of
separation of the boundary layer are highlighted in the second part of the article.
Keywords: Visualizations, PIV measurements, free surface flows, Kelvin-Helmholtz instability,
vortices.
1. Introduction
The flow around a cylinder is a fundamental problem in fluid mechanics with many practical
applications in aerodynamics, hydrology and hydraulics, installations of hydraulic turbines, pipes,
weirs, the equipment using tidal power, submarines, offshore platforms, inflatable dams [1, 2].
The first detailed experimental study of the flow around an immersed cylinder was made by
Sheridan et al. 1997 [3]. The PIV measurements were made for three different cases, namely: (a)
variable Froude number and the ratio constant, (b) Froude number and the ratio
variables, (c) the constant Froude number and the ratio variable.
Hoyt & Sellin 2000 [4] conducted an experimental study on a free surface around a cylinder by
qualitative visualizations, the results being compared with the measurements of Sheridan et al.
1997 [3]. The study was conducted for six values of the ratio , namely 0
0.75, but
only three were analyzed in detail,
0; 0.3; 0.75 for an average velocity of V
0.43 m/s.
Carbery and al. 2004 [5] studied the modification of the position of the separation points of the
boundary layer around the cylinder obtained from the instantaneous vorticity fields and the
variation in time of the lift using the PIV technique for Fr 0.166, Re 2100 and different
values of the ratio
1; 0.125; 0.079.
The aim of this paper is the analysis of a flow around a cylinder by means of PIV measurements
and flow visualizations in order to seize the free surface influences on the flow pattern in the
vicinity of the cylinder and the position of the separation points on the cylinder surface. 2D flow
visualizations, velocity fields and velocity profiles are presented in the interest region of the free
surface and the cylinder in the transitory sub-critic regime with the characteristic Reynolds and
Froude numbers of: 5000
10000, respectively Fr 1.
In addition to the Reynolds number Re
V , the relative position of the free surface and
the interpretation of specific variables in this type of flow, also depend by others parameters, [6]:
V
The Froude number, Fr
, where is the distance between the free surface upstream of

The ratio
the cylinder in the area where there is no disturbance, and of the surface of the cylinder,
see figure 1.
170
, where is the gravitational acceleration;
h
Free surface
V0
Fig. 1 Sketch of the flow around the cylinder
2. Experimental setup
The measurements were performed in a plexiglass channel where a cylinder of diameter
50
mm is immersed. Downstream of the cylinder is a weir which allows the control of the
transported flow rate. The channel has a thickness of 15 mm and length up to weir
613 mm.
For this series of measurements the height at the entrance of channel
(x = 0) and the height of
the weir were kept constant, see figure2.
free surface
cylinder
weir
Fig. 2 Experimental setup
Laser source
PIV camera
Measurement
region
Fig. 3 PIV setup and the measurement region
For the PIV measurements a CMOS Nano Sense MKII camera was used, with a maximum
acquisition rate of 5 kHz and a maximum resolution of 512 512 pixels (see figure 3).
Measurements consisted in the recording of series of pairs of images for frequencies between 1
and 3 kHz. The light source was an infrared laser Nanopower with a power of 4W and a
wavelength of 795nm. The seed particles are glass spheres coated with a silver layer.
Using PIV method whole velocity fields can be measured in a plane by taking series of two
images at a specified time one after another. In this way by the PIV method the magnitude and
the direction of the velocity field can be determined. After the system calibration, the distance
171
traveled by an individual particle can be calculated, determining in this way, from the known
time difference and the measured displacement, the velocity values of each particle in the field.
The principle of this measurement method is the determination of the local flow velocity from
the relative movement of the particles. During a time interval , a group of particles moves
from position to position
, and the local movement velocity is:
,
(1)
Calibration images resulted in a spatial resolution of 0.07mm/pixel which corresponds to a field

of view of 36mm x 29mm. The 350 pairs of images were processed using an adaptive algorithm
on several levels considering the cell deformation and sub-pixel displacement. The final grid is
composed of cells of 16 x 16 pixels with 75% overlapping which can be translated as a vector for
each 4.12 pixels or a spatial resolution of 0.36 mm.
In both cases the correlation method was validated at each level if the SNR correlation value is
higher than a threshold set at 1.1. On average, less than 2% of vectors have been found to be
invalid. They have been corrected by means of a bilinear interpolation. A systematic inspection
of displacement histograms show a good discretization with a bimodal distribution (without the
presence of "peak locking" phenomenon) [3].
The measurements were performed for three immersion heights of the cylinder. The geometric
parameter characteristics are given in Table 1.
Table 1
Sizes obtained experimentally
No.
1
2
3
[mm]
92
105
115
[mm]
6.4
18.76
29.2
[-]
5400
7500
10500
[-]
0.154
0.22
0.3
[-]
0.128
0.375
0.584
3. Flow visualizations
For the discharges occurring at low heights, the free surface has irregularities due to
hydrodynamic instabilities of Kelvin-Helmholtz (see figure 4). Indeed, downstream of the
cylinder there is a hydraulic jump, disrupted by Kelvin-Helmholtz instabilities and the effects of
surface tension, respectively, by the manifestation of the adhesion forces of the solid surface.
We note upstream of the cylinder that the free surface is stationary, which implies the existence
of a continuous flow regime. Downstream of the cylinder, the free surface instabilities show a
non-stationary regime of the flow.
Free surface instability

Midplane
Fig. 4 - Visualizations of the free surface flow and the presence of fluctuations downstream of the cylinder
172
In Figure 5 visualizations spectrum are presented downstream of the cylinder obtained by the two
different ways: one using the same PIV system used for measurements and one by a photography
using a colored blue dye.
a)
b)
c)
Fig. 5 - Flow visualizations using PIV system and a photographic technique for different heights of immersion of
the cylinder
In these images the separation of the fluid from the surface of the cylinder and the appearance of
vortices downstream of it can be observed. For a steady regime upstream the cylinder, we can
highlight the appearance of instabilities downstream the cylinder in the vicinity of the free
surface. The presence of the free surface and the asymmetric position of the cylinder in relation
to the distance between the free surface and the bottom surface of the channel, influences the
formation of a von Krmn vortex sheet with a deformed structure, see figure 6.
Fig. 6- Visualization of a von Krmn vortex sheet downstream the cylinder

173
4. Flow field measurement

PIV measurements have allowed the determination of the mean velocity fields and the precise
identification of the separation point of the boundary layer on the surface of the cylinder and the
determination of the line of the free surface. One hundred velocity profiles have been extracted
from the region of interest for the studied flow (see figure 7).
The profiles represented in figure 8 show the evolution of the free surface characterized by
fluctuations for In > 40 as well as the area near the cylinder where the appearance of a
recirculation is possible downstream the separation point.
Area of interest
l1
l25
l100
Fig. 7 Distribution of velocity vectors in the plane and definition of the region of interest and I1-I100,
I25 profiles corresponds to x 303 mm where is the center of the cylinder
(corresponding view of the Reynolds numberRe 7500)
free surface evolution
fluctuations
recirculation
Fig. 8 Velocity profiles obtained from PIV measurements in the interest region.
174
5. Conclusions
Two different types of visualization techniques were carried out in this paper, one by the same
PIV system used for measurements and one by a photography technique using a colored blue dye,
with the purpose of analyzing the flow around a 50mm cylinder located in a 15mm narrow
channel in the vicinity of the free surface for different immersed depths. The advantage of the
photography technique using colored blue dye is that a larger area of the flow is covered enabling us
to see large flow structures like von Krmn vortex sheet downstream the cylinder. The advantage
of visualization technique taken with the same PIV system used for measurements is that it has a
much better resolution which allows visualization of small flow structures in the flow with much
better precision but in smaller investigation zones than the previous technique.
PIV measurements and direct streak lines visualizations are used to analyze the wake structure.
A von Krmn vortex was observed downstream the cylinder, its generation being influenced by the
presence of the free surface and the asymmetric position of the cylinder in the channel. The
instability that occurs downstream of the cylinder and the separation area of the fluid from the
cylinder were highlighted. The visualizations disclose the presence of Kelvin Helmholtz
instabilities between the free surface and the fluid domain, respectively in the vicinity of the
separation point. From the interest region, 100 of velocity profiles were extracted through PIV
measurements in this way succeeding a complete characterization of the flow field in the
measurement area.
References
[1]. Whelan J. I., Graham J. M. R., Peir J., A free-surface and blockage correction for tidal turbines, J. Fluid
Mech., 624 (2009), pp. 281291.
[1] Myers L.E., Bahaj A.S., Experimental analysis of the flow field around horizontal axis tidal turbines by use of
scale mesh disk rotor simulator, Ocean Engineering, 37 (2010), pp. 218227.
[2] Sheridan J., Lin J-C., Rockwell D., Flow past a cylinder close to a free surface, J. Fluid Mech. 330 (1997), pp.
1-30.
[3] Hoyt J. W., Sellin R. H. J., A comparison of tracer and PIV results in visualizing water flow around a
cylinderclose to the free surface, Exp. Fluids 28 (2000), pp. 261-265.
[4] Carberry J., Sheridan J., Rockwell D., Cylinder oscillations beneath a free-surface, European J. Mechanics B /
Fluids 23 (2004), pp. 8188.
[5] Sheridan J., Lin J-C., D. Rockwell, Metastable states of a cylinder wake adjacent to a free surface, Phys. Fluids
7 (1995), pp. 2099-2101.
175
DYNAMIC BEHAVIOR OF BUCKLING RESTRAINED BRACES AND

THE INFLUENCE OF COMPRESSION STRENGTH
ADJUSTMENT FACTOR
Mircea VADUVA PhD Student, Technical University of Civil Engineering of Bucharest,
e-mail: mircea.vaduva@gmail.com
Lidia MARIN PhD Student, Technical University of Civil Engineering of Bucharest,
e-mail: lidia.mihaela@gmail.com
Abstract: In the last years, the buckling restrained braces (BRBs) have been used more and more
because of their capacity to sustain large inelastic deformations under compression stress, without
buckling effects. This study analyzes the dynamic response of BRBs, modeled in two different
ways. Moreover, the way the compression strength adjustment factor influences the energy
dissipated during dynamic actions is also analysed. In order to obtain the results 12 nonlinear
dynamic analyses were performed, six for buckling restrained braces modeled by using link
elements and another six for the same system modeled by using linear elements and plastic hinges.
The compression strength adjustment factor varied, from the situation where the tension capacity of
the BRB is equal to the compression capacity to the situation where the compression capacity is
40% bigger than the tension capacity. The dynamic input used in this study is a harmonic action
and a recorded accelerogram. The energy dissipated by the systems during the first cycle of the
harmonic action was monitored and the values were compared in all twelve cases. Furthermore, the
total energy dissipated during the seismic action was recorded and some differences were noticed
between the two ways of modeling buckling restrained braces. A few conclusions resulted from
these comparisons and the most important one is how a BRB should be modeled in order to obtain
a seismic response close to experimental results.
Keywords: buckling restrained braces, dissipative system, hysteretic device, nonlinear behavior,
compression strength adjustment factor
1. Introduction
Buckling restrained braces (BRB) have the ability to sustain large inelastic deformations without
significant loss of strength under compression and because of the fact that they have a very good
response under cyclic actions. They were first used in Japan in the 70s, and since then the
system has been improved, a lot of buildings being designed with this kind of braces lately.
This study analyzes the response of the buckling restrained braces subjected to dynamic stresses
and the influence of the compression strength adjustment factor to the energy dissipated during
these actions. The BRBs were modeled in two different ways: first they were modeled using
plastic link elements and then using linear elastic elements with plastic hinges assigned, in order
to achieve the plastic deformation of the brace. The influence of the compression strength
adjustment factor was studied, having the following values: 1.00, 1.10, 1.20, 1.30, 1.35, 1.40. In
all these cases the energy dissipated during dynamic actions was recorded and compared.
2. Structure description
2.1. Geometry
The response of a single BRB was analyzed. The section of the metal core that reaches inelastic
domain is 20 x 20 cm. The length of the brace is 4.0 m. In Fig. 1 two models used in this study
are presented. The buckling restrained brace was considered fixed at both endings and at the
superior one there is an imposed displacement equal to 2.5% from the length of the brace,
meaning 10.00 cm.
176
Fig. 1 Two ways of modeling BRB
2.2. Loads
There are two types of loads used in this analysis: one generated by an earthquake, recorded in
1977 in Bucharest, and a set of harmonic motions. In Fig. 2 the recorded accelerogram used in
this study is presented.
Fig. 2 Vrancea 1977 Accelerogram
The harmonic loads were modeled by raising and lowering their intensity in order to obtain a set
of actions that acts on the buckling restrained braces. This pattern was scaled with these factors:
4%, 10%, 30%, 50%, 70%, 90%, 100%, 110% from the maximum deformation. In Fig. 3 the
periodic load used in this study is presented.
According to [3] the number of test cycles at the design displacement level will be increased for
devices with fundamental periods considerably less than 2s. Structures equipped with BRBF
generally have a fundamental period in between 1-2s, therefore an appropriate increase in load
cycles is necessary. For the considered case, 10 cycles for the maximum amplitude were used
and 5 cycles for 25% and 50% from the maximum amplitude. In Fig. 4 this type of load is
represented.
177
Amplitude
Increasing force
1
0.75
0.5
0.25
0
-0.25 0
10
15
20
-0.5
-0.75
-1
Time
Fig. 3 Harmonic action (SAP2000) [2]
Fig. 4 Increasing force action
3. Structure modeling
3.1. Materials
For the buckling restrained brace the material used in this study was S355, with fy=235N/mm2.
3.2. BRB
The plastic link element used to model the buckling restrained brace is an element that connects
two points, i and j, that are placed at a distance L one from another. Between these two points a
structural behavior for six degrees of freedom, linear or nonlinear can be defined. In Fig. 5 the
Multilinear Plastic element used in this analysis is presented, from the software SAP2000. The
behavior of the buckling restrained braces was modeled in the same way for tension and
compression. Depending on the value of , the compression strength adjustment factor, the
behavior can be symmetrical or not.
Fig. 5 Multilinear plastic link (SAP2000) [2]
Fig. 6 Plastic Hinge (SAP2000) [2]
For the second model, a linear element with an assigned plastic hinge was used. The hinge was
modeled to describe the natural behavior of the BRB. At the beginning of the analysis the
behavior was considered symmetrical and then, after the compression strength adjustment factor
178
was increased, the compression capacity was bigger than tension one. In Fig. 6 the way the
plastic hinge is presented was defined in SAP2000.
Fig. 7 shows the capacity of a buckling restrained brace, which was computed according [4], as
following:
Maximum compression capacity:

,
Maximum tension capacity:
,
Where:
(3.1)
(3.2)
- Compression strength adjustment factor:

the ratio of the maximum compression force
to the maximum tension force of the test
specimen measured from the qualification
tests
Fig. 7 Capacity curve of the BRB
- Tension strength adjustment factor : the ratio of the maximum tension force measured from
the qualification tests
Npl,Rd Yeld forceAccording to [1], experiments done on the buckling restrained braces showed
values for these two correction factors. For the compression strength adjustment factor, , the
maximum value obtained in the tests was 1.40, and for the tension strength adjustment factor
1.35was obtained.
Because these factors can vary depending on the manufacture of the BRB, in this study the values
for the compression strength adjustment factor were taken in order to see its influence to the
behavior of the buckling restrained braces. In this study, the tension strength adjustment factor was
considered 1.35.
For the last case analyzed the capacity of the buckling restrained braces was:
ov=1.0
Ty=200*200*235*10-3=9400KN
Cy=Ty=9400KN
Tmax=1.35*9400=12690KN
Cmax=1.35*1.40*9400=17766KN
4. Performance assessment
The BRB was modeled in two different ways and the compression strength adjustment factor
varied so that six cases resulted. As the results are typically the same in all six cases, further in
this article only the results from the last case will be presented, where = 1.35 and = 1.40.
First, for the multilinear link solution, in Fig. 8 the hysteretic curves obtained during the dynamic
analysis with an imposed displacement equal to 4% of the maximum allowed displacement are
presented. The buckling restrained brace remains in the elastic domain and all five cycles of motions
are identical, the curve looking like a straight line. In Fig. 9 the hysteretic curve obtained at 90% of
the maximum allowed displacement is presented. It should be observed that BRB is leaving the
elastic domain and has stable plastic deformations in all cycles of action.
In Figure 10 the result for the 4% case is presented when the brace is modeled as a linear
element. Like in the other cases, BRB remains in the elastic domain of deformation. However,
when we look at the hysteretic curve obtained for the 90% case, presented in Figure 11, we can
see that there is a difference between the two ways of modeling the BRBs: stiffness modifies
itself when the action changes its sign.
179
4%
90%
5000
0
-1E-17
-5000
-0.005
20000
Force [KN]
Force [KN]
10000
10000
0
-0.1
0.005
Fig. 9 Link, 90%
4%
90%
Force [KN]
5000
0
-1E-17
-5000
-0.1
0.005
0.1
90%
Fig. 8 Link, 4%
10000
Force [KN]
0.05
-20000
Displacement [m]
-10000
Displacement [m]
-0.005
-0.05
0
-10000
10000
0
-0.05
0
-10000
0.05
-10000
Displacement [m]
-20000
Displacement [m]
Fig. 10 Hinge, 4%
Fig. 11 Hinge, 90%
0.1
90%
The difference between the two models in terms of energy dissipated is the following: for a
single cycle of action the energy dissipated by the linear element with plastic hinge is bigger than
the multilinear plastic link with about 17%, like in Fig. 12 and 13. In the other cases the
difference remains around 18%.
The influence of the compression strength adjustment factor is very small. In case 5 its
value was 1.35 and in case 6 it was 1.40. For this improvement the difference between the
energy dissipated in both cases in the first cycle of harmonic action is 0.01%, so the
influence is almost zero.
When the two models were subjected to seismic motion a bigger difference between the energy
dissipated by the braces was observed. In Fig. 14 the total energy dissipated can be observed.
The difference in this case is around 28%, but still like in the other cases, when the compression
strength adjustment factor was smaller.
0.15
10000
4000.00
0
0.1
0.05
0
10000
3,857.64
3289.71
3000.00
0.05
0.1
20000
Displacement[m]
Fig. 12 First hysteretic curve
180
Dissipatedenergyinfirstcycle
5000.00
Energy[KNm]
Force[KN]
Onehystereticcurve
0.15
link
2000.00
1000.00
0.00
Hinge
Link
Fig. 13 Difference between energy dissipated in the first

cycle
Fig. 14 Total energy dissipated during seismic action
The same problem with the stiffness can be observed if we look at the Fig. 15 and Fig. 16, where the
hysteretic curves obtained during the increasing force action are plotted. Also in Fig. 17 and Fig. 18
all curves that resulted from the analysis with the increasing force are presented.
20000
0.2
0.1
0.1
4%
10
%
30
%
50
%
70
%
0
0.2
20000
Displacement[m]
Hystereticcurvesforlinearelement
withplastichinge
Force[KN]
Force[KN]
Hystereticcurvesformultilinear
link
20000
4%
10000
10%
0
0.2
0.1
10000 0
0.1
30%
0.2
50%
20000
Displacement[m]
70%
Increasing force results for brace

modeled as multilinear plastic link
Increasing force results for brace

modeled as linear element with
20000 hinge
plastic
-0.15
-0.1
15000
10000
5000
0
-5000 0
-0.05
0.05
-10000
-15000
-20000
Displacement [m]
Force [KN]
Fig. 16 Hysteretic curves for linear element with

plastic hinge
Force [KN]
Fig. 15 Hysteretic curves for multilinear link
0.1
0.15
-0.15
Fig. 17 Link, Increasing force
10000
0
-0.1
-0.05
0
-10000
0.05
0.1
0.15
-20000
Displacement [m]
Fig. 18 Hinge, Increasing force
5. Concluding remarks
In the case when the BRB is modeled by using the multilinear plastic link, the braces had a stable
behavior under the action of dynamic forces. The hysteretic curves that were obtained show that
the energy dissipated is very close to the theoretical one.
When the BRB is modeled by using the linear element and the plastic hinge attached, the
buckling restrained braces have a disadvantage when the brace is unloaded, the stiffness not
being the one from the elastic domain, which is not correct.
181
Energydisipationforharmonic
action
6
5
4
3
2
1
Case
Case
The energies dissipated in the first cycle of the action are presented in Fig. 19 and in Fig. 20 the
total energy dissipated during the seismic motion is presented.
Hinge
link
0.00
1000.00
2000.00
3000.00
6
5
4
3
2
1
4000.00
Energydisipationforseismic
motion
Hinge
link
0.00
10000.00
Energy
20000.00
30000.00
Energy
Fig. 19 Energy dissipated for harmonic action
Fig. 20 Energy dissipated for seismic action
After studying the influence of the compression strength adjustment factor on the energy,
dissipation it was observed that in all six cases this influence is very small. This can be explained
by the fact that the deformation of the buckling restrained braces is limited by the maximum
deformation of the system. In all six cases, when the systems were subjected to harmonic
actions, the difference between the energy dissipated in the first cycle of action remains around
18%, like in Fig. 21. The exception is the case when the compression capacity of the BRB was
equal with the tension capacity, which led to a difference of 15%.
Differences between the results

obtained for seismic action
Case
Case
Differencesbetweentheresults
obtainedforharmonicaction
5
3
0.00%3.00%6.00%9.00%12.00%
15.00%
18.00%
0.00%
Differencesbetweenenergydissipatedin
thefirstcycle
Fig. 21 Differences between energy dissipated for
harmonic action
10.00%
20.00%
30.00%
Differences between total energy

dissipated
Fig. 22 - Differences between energy dissipated for
seismic action
In all six cases when the BRB was subjected to sesimic motion, the influence of the compression
strength adjustment factor on the energy dissipation was also very small. However, in this case, the
difference between the total energy dissipated during the earthquake was bigger than in the case of
harmonic actions. This can be explained due to the high number of cycles of loadings and
unloadings that can be found in the seismic action and the error from the model with linear element
and plastic hinge. This error increased after each cycle. The differences are presented in Fig. 22.
References
[1] EN 15129 "Anti seismic devices", 2009
[2] CSI, SAP2000 Basic Analysis Reference Manual, 2009
[3] L. Dunai, "Type testing of buckling restrained braces according to EN 15129," 2011
[4] Romanian seismic code P100-1/2011 (Cod de proiectare seismica P100-/2011)
182

Scientific Journal - Special Issue - November 2013

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Scientific Journal - Special Issue - November 2013

Transféré par

Droits d'auteur :

Formats disponibles

SCIENTIFIC JOURNAL OF THE TECHNICAL UNIVERSITY

MATHEMATICAL MODELLING OF SHOCK WAVES GENERATED BY BLAST EVENTS AND THEIR

Mathematical Modelling in Civil Engineering -Special Issue - 2013

CONSOLIDATION PROCEDURE FOR UNSATURATED SOILS USING A

Fig. 1 - Constitutive surface of an unsaturated soil [1]

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Fig. 2 The proposed unsaturated single cell triaxial setup

Fig. 3 - The flushing device according to Fredlund [1]

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Mathematical Modelling in Civil Engineering -Special Issue - 2013

SLOPE STABILITY ANALYSIS OF UNSATURATED SOILS

The strength parameters are independent of stress-strain behavior;

Forces involved in the equilibrium methods are statically indeterminate.

2.1. Stress Analysis Method (SAM)

is the shear stress, and

is the shear strength according to Mohr-Coulomb failure

Fig. 1 - Stress components of sliding surface [4]

Where, c is the effective cohesion,

(the total unit weight) and

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Occurrence of large deformations;

The effect of three dimensional conditions;

Occurrence of narrow, continuous localized zone on which the slope slides.

Fig. 2 - Total stress analysis of the slope [4]

2.2. Strength Reduction Method (SRM)

Mathematical Modelling in Civil Engineering - Special Issue - 2013

Fig. 3 - Strength Reduction Method [6]

is the shear strength of the sloped material.

This is computed according to the Mohr-Coulomb criterion.

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Soil parameters for case 1

Soil parameters for case 2

Fig. 4 - Geometry and finite element mesh

3.2. Numerical Results for First Case

Fig. 5 - a) Displacements contour of the slope; b) Shear strength of the slope.

2D results using SRM method without water table. FOS = 1,30.

Fig. 6 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 7 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 8 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 9 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 10 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 11 - a) Displacements contour of the slope; b) Shear strength of the slope.

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Fig. 12 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 13 - a) Displacements contour of the slope; b) Shear strength of the slope.

Fig. 14 - a) Displacements contour of the slope; b) Shear strength of the slope.

Mathematical Modelling in Civil Engineering - Special Issue - 2013

Numerical result for slope analysis soil 2

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Mathematical Modelling in Civil Engineering - Special Issue - 2013

Fig. 25 - Comparison of FOS on SAM and SRM methods

Mathematical Modelling in Civil Engineering -Special Issue - 2013

EXPERIMENTAL AND NUMERICAL INVESTIGATION OF INDOOR

Fig.1 - Photos of the TESTO 480 installed in the classroom

Mathematical Modelling in Civil Engineering -Special Issue - 2013

CO2 level [ppm]

Illuminance level [lux]

Temp level [oC]

Fig.3 - Annual heating, cooling and electric energy consumption

Cooling demand for the hottest day of the school year

Mathematical Modelling in Civil Engineering -Special Issue - 2013

Operative temperature [oC] Coldest day

Operative temperature [oC] Warmest day

Mathematical Modelling in Civil Engineering - Special Issue - 2013