Vous êtes sur la page 1sur 19

The 2011 World Congress on Advances in Structural Engineering and Mechanics (ASEM'11+) Seoul, Korea, 18-22 September, 2011

Fuzzy-Monte Carlo simulation for the safety assessment of bridge piers in presence of scouring Pier Giorgio Malerba1) Elsa Garavaglia2) Luca Sgambi3)
1), 2), 3)

Department of Structural Engineering, Politecnico di Milano, Milan, Italy 1) malerba@stru.polimi.it

ABSTRACT Bridge scour is the removal of sediment (sand and rocks) around bridge abutments and piers due to the different river flow over the year. Approximately the 60% of collapses occurring to the of bridges is directly or indirectly associated to scour; this phenomenon is very important in the bridges safety assessment. The depth of scour depends on many factors: type and amount of bed/embanked material transported by the river flow, nature and occurrence of flood, accumulations of debris, constriction or realignment of flow due to the presence of the bridge, geometry and orientation of piers, etc. Under a design point of view, the depth of scour can be evaluated through theoretical models involving hydraulic parameters. However, many of these parameters (flow characteristics, sediments, structural and geotechnical factors, etc.) are affected by uncertainty; in the safety assessment of bridges, a deterministic approach to the problem cannot be adopted. A proper assessment of structural safety must be performed taking into account the uncertainties involved in the problem. The uncertainty can derive by randomness so as by incomplete and inaccurate datasets and the development of a fuzzy-probabilistic approach permits to take into account some uncertainty involved into the problem. This paper proposes and discusses a Fuzzy Monte Carlo Simulation (FMC) to approach the assessment of the bridge safety in presence of scouring. To verify the feasibility of the proposed approach the FMC will be applied to the safety assessment of a case study.

1. INTRODUCTION In the reliability assessment of bridge piers in riverbed, a primary role is played by the scouring action. A study reported by Hamill (1999) regarding the collapse of 143 bridges between 1847 and 1975 shows, in fact, that 66 bridges collapsed for problems associated to the scouring effect. More recent analyses, concerning the vulnerability of bridges in the Po Valley, show that the most common bridge damage is caused by erosion occurring in riverbed foundations (Ballio et al. 1999).
1) 2) 3)

Full Professor Associate Professor Assistant Professor

A safe design of bridges with piers in riverbed cannot be carried out without a careful study of the scouring effects. In this context, the hydraulics studies the flow and its transition characteristics through the bridge waterway and leads to define the drag forces and the scour depths (Hamill 1999, FHA 2001), the static firstly work to the design of pile and foundations and then checks their bearing capacity under the most severe loading and riverbed conditions. Unfortunately, the design computations are not straightforward and they intrinsically involves a great deal of parameters, many of them strongly affected by uncertainties, like irregularity of the channel geometry, variations in water flow, velocity in the contraction zone, shape, dimensions and orientation of obstacles, presence of debris and floating objects.

Fig. 1. Bridge pier with debris. Some of these uncertainties and their influence on the structural response were examined by the authors in a previous work (Dordoni et al. 2010), where the uncertainties were modeled using fuzzy theories only. In the present paper, recognizing the existence of different types of uncertainties, a fuzzy-random approach will be developed. 2. CONVENTIONAL ASSESSMENT OF HYDRAULIC FORCE AND SCOURING The safety assessment of a bridge must take into account the problems caused by the erosive action of the river stream (Da Deppo et al. 2006, Muzzamil et al. 2008). Generally speaking, the reason of this phenomenon is to be found in the imbalance between the sediment transport capacity of the stream and the actual amount of suspended solids in the fluid, which depends

also on the availability of solid particles along the bed to be carried by the fluid motion. In these terms, erosion and deposition of material are opposite manifestations of the same phenomenon. Piers, abutments and embankments of river bridges can exert a contraction on the flow with respect to its natural conditions, because they occupy a space that would be otherwise available to the stream. In many situations, when the undisturbed flow is slow, the local contraction of the channel width can increase the velocity of the flow, exposing the material of the riverbed in the proximity of bridge piers (and abutments) to the erosive action. 2.1. Hydraulic force The hydraulic force, exchanged between water streams and solid bodies in relative motion, can be estimated through Eq. 1 (Apelt and Isaacs 1968, Blevins 1984):
FD = 1 V22 y2 L C D 2

(1)

where and V2 are the density and the average flow velocity, y2 is the average water depth at the contraction and L is the pier dimension in the direction of the flow. The drag coefficient CD depends on the shape of the piers, the distance between them, the attack angle of the flow and the Reynolds number, which is proportional to the ratio between inertial forces and viscous forces in the fluid motion. The drag coefficient can be estimated experimentally. Unfortunately, in literature, values are available only for profiles with simple geometric shapes; in case of actual bridge piers, the drag coefficient can vary between 0.2 and 2.0 (Blevins 1984). This wide range of variability provides high uncertainty in the estimation of the drag coefficient and force. 2.2. Scouring assessment For the assessment of the scour depth, a common engineering practice suggests that it can be evaluated as the sum of depths due to the three levels in which the phenomenon occurs: contraction scour, local scour and degradation-aggradation of the riverbed (Hamill 1999, U.S. Dept. of Transp. 2001). 2.2.1 Contraction scour Contraction scour (dsc) is a consequence of the reduction of the available width to the flow due to the presence of the bridge support. With reference to Fig. 5 and 6, let be respectively y1 and b1 the average water depth and the width of the main channel upstream of the contraction, Q1 the discharge portion that carries solid particles upstream of the contraction, Q2 e b2 the total discharge passing through the waterway and the waterway width, k1 a coefficient that depends on the amount of sediment present in the channel. On the basis of these quantities, when the undisturbed flow upstream of the contraction contains suspended solid particles, i.e. the case of live-bed conditions, the scour depth can be assessed through the following Eq. 2 and 3 (Laursen 1962, Richardson et al. 2001). Q b y = y1 2 1 Q1 b2
' 2 6 7 k1

(2)

' d sc = y2 y1

(3)

2.2.2. Local scour Local scour occurs when the flow collides with localized obstacles, such as supports and piers of the bridge. The velocity field of the stream is abruptly distorted, generating vortices and vertical motions that remove material from around the bottom of the obstacle. Calling F2 the Froude number relative to the water flow at the beginning of the restriction, K1P, K2P, K3P corrective factors that depend on the shape of the piers and of the river bed and bp the width of the piers, the depth of local scour (dsp) can be estimated using Eq. 4 and 5 (Hamill 1999): F2 = V2 g y2
0.65

(4)
F20.43

b d sp = 2 y2 K1P K 2 P K 3 P p y2

(5)

2.2.3. Degradation and aggradation Degradation and aggradation in the fluvial bed can be due to changes in the river basin, climate trends, or the amount of solid material carried by the flow (Miglio et al. 2009, Goode and Burbank 2009). This process occurs on larger time and space scales respect to the two previous phenomena. In this paper degradation and aggradation of the fluvial bed is not considered, and the total depth of erosion (ds) can be approximately assessed as the sum of the first two effects (contraction and localized erosion):
d s = d sc + d sp

(6)

From what has been previously stated, it is clear that the hydraulic force and scour depth depend from the many uncertainties involved in the problem. Deterministic analysis, based on the Eq. 16, may be inaccurate, since it does not take into account the uncertainties related to the factors present in the previous equations. A more reliable assessment can be performed using an analytical model that takes into account all types of uncertainties involved in the problem. 3. TAXONOMY OF UNCERTAINTY Unfortunately, many parameters present in the Eq. 1-5 are affected by uncertainty. The average flow velocity V2, for example, is variable over time and space and cannot be considered a constant parameter on whole model. The values of corrective factors K1P, K2P, K3P are based on experimental evidence, therefore they are affected by uncertainty correlated to the process of extrapolation data.

In common sense, uncertainty refers to a lack of data or confidence about something (Elith et al. 2002). Clearly, there are several sources of uncertainty in engineering practice and different methods to handle each of them. Before describing the uncertainties considered in this paper, it is useful to propose a more general classification of uncertainties in order to understand the motivations of the approach used in the following paragraphs. Reagan et al. (2002) propose a classification of uncertainties, involved in the biological field, which can be conveniently applied to other fields of science. They classify the uncertainties into two main groups: epistemic and linguistic. Epistemic uncertainty is associated with knowledge of the state of a system and it includes uncertainty due to limitations of measurement devices, insufficient data, variability over time or space, etc. Linguistic uncertainty arises because much of our natural language is ambiguous, vague, context dependent. The uncertainties can be classified as Fig. 2 (Reagan et al. 2002):

Taxonomy
Measurement error Systematic error Natural variation Epistemic Inherent randomness Model uncertainty Subjective judgment Uncertainty Vagueness Context dependence Liguistic Ambiguity Underspecicity Indeterminacy of theoretical terms

Treatments
Probability

Probability Probability Fuzzy

Fuzzy

Fig. 2. A taxonomy and a treatment of uncertainty. For each type of uncertainty there are several methods for its treatment in numerical models. In literature are presents methods based on probability theory, statistical techniques, analytic

error estimation, methodologies to recognize and remove bias, fuzzy theory, fuzzy sets, threevalued, rough sets, etc. In the present paper some of the uncertainties classified in Fig. 2 are modelled by probabilistic and fuzzy approach. In a probabilistic approach, the variables are random variables (RV) while in a fuzzy approach, the variable are named fuzzy variable (FV). In the following paragraphs, random and fuzzy uncertainties are described briefly. 3.1 Random variables A random variable (RV) is a variable whose value results from a measurement on some type of random process. A real random variable X ( ) , with , is a function mapping a sample

experiment such that the set { X ( ) x} is an event for any number x and the probabilities

space into the real line R. The real number X ( ) corresponds to every outcome of an

P { X ( ) = +} = P { X ( ) = } are zero (Elishakoff 1983).

associated cumulative distribution function FX ( x ) . In this paper random variables are considered depending from a normal probability distribution. Eq. 7-8 are, respectively, the probability density function and the cumulative distribution function for the Normal distribution.

A real random variable can be described by its probability density function f X ( x ) and the

1 x a 2 fX ( x) = exp with < x < and X > 0 X 2 2 X 1 1 x a 2 FX ( x ) = exp dx X 2 2 X 1


x

(7)

(8)

In Eq. 7- 8 the parameters a and X represent the mean value and the standard deviation of the distribution. Fig. 3 shows graphically these functions.

Fig. 3. Example of a random variable (Normal distribution): probability density function (left) and cumulative distribution function (right). Random variable can be used to model the uncertainty associated with random event. To handle random uncertainties, Monte Carlo simulation is very effective. This method represents

an important technique in computational science. In a Monte Carlo simulation, through a large number of numerical simulations of the deterministic problem, it is possible to get statistical information about the behaviour of the model having uncertain data. Usually, Monte Carlo methods consist of three steps: Step 1: Simulation of the random variable function. Step 2: Solution of the deterministic problem for a large number of realizations. Step 3: Statistical analysis of the results. Fig. 4 shows schematically this process.

Fig. 4. Schematic illustration of Monte Carlo method.

3.2 Fuzzy variables A fuzzy variable (FV) in fuzzy theory is represented by fuzzy sets (Zadeh 1965, Zadeh 1968, Piegat 2001). A fuzzy set A is a set of pair: A=

{(

( x ) , x )}

(9)

where: A is the membership function which assigns to each element x the grade of its membership to the fuzzy set A:

A ( x ) [ 0, 1]

(10)

The membership function maps the numerical universe X of a given variable in the interval [0,1]. In the practice, various types of membership function are applied (triangular, rectangular, trapezoidal, Gaussian, sigmoid, etc.). Fig. 5 shows a fuzzy variable having parabolic shape.

Fig. 5. Characteristic parameters of a fuzzy set.

Fig. 6. Mapping of a fuzzy input x to a fuzzy output y.

of fuzzy output variable. Let y = f ( x ) a mapping function that links the input variable to output function y = f ( x ) and a series of level cuts, it is possible to determine the shape of the variable. Then, using the membership function of the input variable A ( x ) , the mapping

Let A ( x ) a membership function of fuzzy input variable and B ( y ) a membership function

membership function of the output variable B ( y ) (Biondini et al. 2004, Sgambi 2004). Fig. 6 shows schematically this process. At last, to obtain a single value, associated with the output variable, the membership function B ( y ) have to be defuzzified through a defuzzification method (Amaya et al. 2009).

3.3 Fuzzy-random variables Random variables are used to handle random uncertainty, fuzzy variables to handle vagueness or not random uncertainty (Fig. 2). However, it is possible that a variable present both types of uncertainty. In this case, the use of a random variable or a fuzzy variable is in any case inadequate to fully describe the involved uncertainty. Kwakeraank (Kwakeernak 1978, Kwakeernak 1979) introduced the concept of fuzzy-random variable to govern both the random and the fuzzy uncertainty on the same variable. The fuzzyrandom variable (Fig. 7) was introduced as a natural generalization of random variable in order to represent relationships between the outcomes of a random experiment and nonstatistical inexact data. Different researchers have studied fuzzy-random variable both from a purely mathematical aspect (Puri and Ralescu 1986, Zadeh 2002, Li and Liu 2006) and in applicative terms (Moller 2002, Kala 2007, Sobhani and Ramezanianpour 2009, Holicky 2010).

Fig. 7. Sampling of one-dimensional fuzzy random variable. To use a fuzzy-random variable, the definition of the probability given in 3.1 should be

reviewed in terms of fuzzy probabilities. The fuzzy probability P ( Ai ) is the set of all probability P X Ai whit the corresponding membership values P X Ai into account all states of the (also partial) occurrence of X Ai .

( (

)) , which takes

% P ( Ai ) =

{( P ( A ) ; ( P ( A )))
i i

P ( Ai ) = P l ( Ai ) ; P r ( Ai ) ; ( P ( Ai ) ) = ( 0;1]

(11)

More details about the definitions of the fuzzy-random variable can be found in Mller (2002). Using the definition of the fuzzy-random variable and the cumulative distribution of the variables, it is possible to define a sampling procedure. A fuzzy-random variable is defined by a number of points that describe its shape. For a triangular membership function, three points are needed. In the following application parabolic membership functions are used. A parabolic curve is completely defined by the passage for three points. From the Fig. 8 the coordinates of these points are ( x1;1) ; ( x0,5 s ;0.5) ; ( x0 s ;0 ) for the first parabolic curve and

{( x ;1) ; ( x
1

0,5 d

;0.5 ) ; ( x0 d ;0 ) for the second one.

Fig. 8. Definition of a fuzzy-random sample. The ordinate of the points are the values of sureness of the membership functions and are fixed

for each fuzzy-random variable. The position and the size of the support and cannot be completely random but derive by shape of the cumulative distribution functions previously defined. A sampling rule can be built following these 6 steps (Fig. 8): Step 1: The five cumulative distribution functions are built starting form a uniform distribution defined on the users data interval. Step 2: The x1 value is extract from the cumulative distribution function 3 using a sampling rule involving the whole cumulative distribution function. s d Step 3: The x0,5 and x0,5 are extract from the cumulative distribution functions 2 and 4 using a sampling rule on the interval AB . s d Step 4: The x0 and x0 are extract from the cumulative distribution functions 1 and 5 using a sampling rule on the interval CD . Step 5: Control on the acceptable shape of the membership function. Step 6: Return to step 1 until it is defined the desired number of fuzzy-random variables. 4. AN APPLICATION A fuzzy-random assessment of the hydraulic force and of the scour depth on a typical fluvial bridge is performed. The bridge and the river geometries are shown in Fig. 9 (Da Deppo and Datei 1994). The riverbed is considered as a wide main channel of rectangular section 64 m wide, 1 m deep. Two floodplain extend from the sides of the main channel with a slight slope toward the center. The bridge crosses the river in perpendicular direction. The supports and the bridge embankments completely obstruct the floodplains and three 2m-wide piers cause a further restriction of the main channel. The distance among the piers is 16 m and the contracted section is 58 m wide. The dimension of the piers in the other in-plan direction is 12 m. The foundation piles positioning is schematically showed in Fig. 10. Assuming a return period of 200 years, hydraulic evaluations led to asses an average water depth of 2.7 m before restriction and 2.5 m at the restriction. The average velocity of the stream in the restricted section is assessed to be 2.14 m/s while the discharge in the main channel upstream of the bridge is assessed to be 280 m3/s. To understand the influence of the uncertainties in the considered problem, the hydraulic force and the scour depth are assessed, both in the case of pier shaft exposed to the stream (Case A) and also in the case of foundation piles exposed to the stream after a possible scour process (Case B). The scour profile along the cross section has not been calculated: only the maximum depth at the bottom of the piers was considered (as representative of the riverbed configuration after scour in the most vulnerable locations).

16 m

16 m

16 m

16 m

b P =2m

y2 = 2.5 m

2.5 m

b 1 = 64 m L = 12 m

160 m

64 m 1m y1 = 2.7 m

160 m
y1 =2.7 m

y2 =2.5 m

b1 = 64 m

Fig. 9. Schematic representation of the fluvial crossing: in-plan and cross-section views.
1m
2m 2.7

R0.4

bP max bP min

1.6 m 2.2 m 2.2 m 2.2 m 2.2 m 4.4 m 4.4 m 14 m 4.4 m

y'
2

1.5 m

Fig. 10. Particular of the foundation piles positioning. A Fuzzy-Monte Carlo analysis (FMC) was performed, in accordance with the definition of random and fuzzy-random variables given in the 3 and Tab. 1 and 2. To understand the importance to model the random uncertainty in relation to the vagueness uncertainty modeled by membership function, a Monte Carlo analysis (MC) with only random uncertainty was performed and the result are compared whit the more general Fuzzy-Monte Carlo approach. Deterministic analysis was also performed considering the mean value of the probability density function of the input variable.

3.2 m

4.1 Case A) Pier shaft exposed to the stream In the Case A) the assessment of the hydraulic force and the scour dept is performed considering only the pier shaft exposed to the stream. The assessment of the hydraulic force is provided by Eq. 1, considering the uncertainty involved in the definition of some parameters while the total scour depth is assessed according to Eq. 2 6 as the sum of contraction and local scour. Tab. 1 summarizes the parameters of the probability distributions and membership functions involved in the analysis. In accordance with the 3 a parabolic shape of membership function was used. The support variability was defined as the distance of the vertex of the parabola with respect to the maximum and minimum value of the variable. For each variable a deterministic value (mean of the probability density function) was defined. Note that the mean value and the support variability of the parameters CD, L, bp (drag coefficient, length of the pier, width of the pier) are set taking into account the possibility of accumulation of debris carried by the stream on the upstream face of the pier, and considering a skew angle = 30 between stream and pier alignment. Table 1. Parameters of the probability distributions and membership functions. Variable
Y2

Type Deterministic Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Deterministic Deterministic Deterministic Fuzzy-Random Fuzzy-Random Deterministic Random Random Deterministic Fuzzy-Random

Mean 1.005 t/ m3 2.50 m 2.14 m/sec 1.74 11 m 2.70 m 280 m3/sec 360 m3/sec 64 m 58 m 0.64 2.50 m 2.14 m/sec 1 2.2 1.1 9.806 m/sec2 2.6 m

V2 CD L Y1 Q1 Q2 b1 b2 K1 Y2 V2 K 1P K 2P K 3P g bP

Contraction scour

Standard deviation --0.18 m 0.15 m/sec 0.30 2.2 m 0.18 m 24 m3/sec 30 m3/sec ------0.18 m 0.15 m/sec --0.20 0.06 --0.25 m

Support variability Min Max ----- 10% + 10 % - 10% + 10 % - 30% + 30 % - 30% + 30 % - 10% + 10 % - 20% + 20% - 20% + 20% ------- 10% - 10% --------- 20% ------+ 10 % + 10 % --------+ 20%

Local scour

Drag force

The K2P coefficient is related to the skew angle between the pier and the stream alignment; it was determined according to the method reported by Hamill (1999) and its variability was based on geometry considerations and a variability of 10 % of = 30 . K3P was given a small variability, applied to a common-use value for similar cases. For both of these coefficients, only random uncertainty was considered. Fig. 11 and 12 show the histograms of the drag force and scour dept arising from Monte Carlo and Fuzzy-Monte Carlo analysis with 10 miles sampling. The deterministic values are 111 kN for the drag force and 11.0 m for the scour dept, respectively.
400 400

300

300

Frequency

200

Frequency
0 50 100 150 200 250 300 350 400

200

100

100

0 0 50 100 150 200 250 300 350 400

Drag force [kN]

Drag force [kN]

Fig. 11. Drag force, histograms of Monte Carlo (A) and Fuzzy-Monte Carlo (B).
450 400 350 450 400 350

Frequency

250 200 150 100 50 0 0 3 5 8 10 13 15 18 20 23 25

Frequency

300

300 250 200 150 100 50 0 0 3 5 8 10 13 15 18 20 23 25

Scour dept [m]

Scour dept [m]

Fig. 12. Scour dept, histograms of Monte Carlo (A) and Fuzzy-Monte Carlo (B).

4.1 Case B) Foundation piles exposed to the stream If the scour exposes the foundation piles, the hydraulic resistance will depend on the new geometry of the stream obstacle. The geometrical characteristics of the mesh of the foundation piles are shown in Fig. 10. To simplify the hydraulic analysis, it was assumed that the flow collides only with the piles, excluding the foundation plinth from the analysis. Tab. 2 summarizes the features of the density probability function and the variability of the membership function considered. When the foundation piles are exposed to the stream, the obstacle changes the geometry. In this case it is not simple to describe how the local conditions of motion vary over the time and with respect to the initial conditions.

Some authors suggest considering only the geometry of the piles when the scour is caused primarily by the contraction effect, and only the geometry of the piers when the local scour is stronger. When piles position describes a grid, piles are particularly effective in capturing debris and floating materials. Such configuration will create a single obstacle that will be larger than the sum of the net dimensions of the piles. This new configuration of the obstacle locally changes the conditions of motion. For this reason, the uncertainties involved in the water height ( y2 ) and in the flow speed ( V2 ) are characterized by a greater spread than in the previous case, since further hydraulic computations could not be performed. Table 2. parameters of the probability distributions and membership functions. Variable
Y2

Type Deterministic Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Fuzzy-Random Deterministic Deterministic Deterministic Fuzzy-Random Fuzzy-Random Deterministic Random Random Deterministic Fuzzy-Random

Mean 1.005 t/ m3 2.50 m 2.14 m/sec 1.86 13 m 2.70 m 280 m3/sec 360 m3/sec 64 m 58 m 0.64 2.50 m 2.14 m/sec 1 2.2 1.1 9.806 m/sec2 5.2 m

Drag force

Standard deviation --0.18 m 0.30 m/sec 0.30 2.4 m 0.18 m 24 m3/sec 30 m3/sec ------0.18 m 0.15 m/sec --0.20 0.06 --0.50 m

Support variability Min Max ----- 20% + 20 % - 20% - 30% - 30% - 10% - 20% - 20% ------- 20% - 20% --------- 20% + 20 % + 30 % + 30 % + 10 % + 20% + 20% ------+ 20 % + 20 % --------+ 20%

V2 CD L Y1 Q1 Q2 b1 b2 K1 Y2 V2 K 1P K 2P K 3P g bP

Local scour

Contraction scour

Fig. 13 and 14 show the histograms of the drag force and scour dept arising from Monte Carlo and Fuzzy-Monte Carlo analysis with 10 miles sampling. The deterministic values is 138 kN for the drag force and 14.5 m for the scour dept. Considering the histograms of Fig. 11 and 12, the exposure of the foundation to the stream increases the uncertainty involved in the analysis. As can be seen from the results (Fig. 11 14), the uncertainty involved in the assessment of the drag force is greater than the uncertainty involved in the evaluation of the depth of scour. In

addition, in this the first case, the vagueness uncertainty covers a significant role and a fuzzyrandom approach is suitable to govern both random and vagueness uncertainty.
400 400

300

300

Frequency

200

Frequency
0 50 100 150 200 250 300 350 400

200

100

100

0 0 50 100 150 200 250 300 350 400

Drag force [kN]

Drag force [kN]

Fig. 13. Drag force, histograms of Monte Carlo (A) and Fuzzy-Monte Carlo (B).
450 400 350 450 400 350

Frequency

250 200 150 100 50 0 0 3 5 8 10 13 15 18 20 23 25

Frequency

300

300 250 200 150 100 50 0 0 3 5 8 10 13 15 18 20 23 25

Scour dept [m]

Scour dept [m]

Fig. 14. Scour dept, histograms of Monte Carlo (A) and Fuzzy-Monte Carlo (B). CONCLUSION This paper deals whit the assessment of the hydraulic force and scour dept in bridges with piers in the riverbed. The conventional design uses equations which involve a certain number of uncertainties regarding flow, sediments, structural and geotechnical parameters. By taking into account these uncertainties through a fuzzy-random approach, the obtained results show how a deterministic approach can disregard variations that make the output more severe with respect to structural safety. After introducing the selected models to describe the problem, the probability and fuzzy theory is used to assess hydraulic force and pier scouring in a bridge with piers in the riverbed. The results of the cases studied (pier shaft exposed and foundation piles exposed) are summarized in Fig. 14 and Tab. 3. Fig. 14 show the comparison between random (MC) and fuzzy random (FMC) analysis for all the variables studied (only a boundary shape of the histograms is showed). It is evident that the shape of the histograms is different in the drag force assessment and similar in the scour dept assessment. In all cases the FMC is more severe than

the MC analysis. In conclusion, handle uncertainty on the safety assessment of bridge piers in presence of scouring is very important. To assess a reliable drag force, both random and vagueness uncertainty should be considered.

Fig. 14. Comparison between random (MC) and fuzzy-random approach (FMC) in the assessment of drag force and scour dept. Tab. 3 shows the numerical comparison among deterministic, random and fuzzy-random approach. For the random and fuzzy random analysis, a characteristic value (95% fractile) of the result variable is reported both for Case A (pier shaft exposed to the stream) and Case B (foundation piles exposed to the stream). Concerning foundation calculations, the uncertainty about hydraulic load values has a relatively smaller effect than the combination of other parameters, in particular the geotechnical ones. However, in cases like this, the uncertainty concerning scour depth suggests to opt for deep foundation systems. The setting adopted in this work and the different steps and criteria that were followed, could be suitably used for the construction of expert systems in order to evaluate the reliability of bridges and quantify indicators that could help in considering structures during their service life with a proper attention.

Tab. 3. Quantitative values of drag force and scour dept for Case A (pier shaft exposed to the stream) and Case B (foundation piles exposed to the stream). Deterministic analysis 111 11.0 138 14.5 Random analysis 154 (+39%) 11.5 (+5%) 228 (+65%) 17.6 (+21%) Fuzzy-Random analysis 162 (+46%) 11.6 (+5%) 252 (+83%) 17.8 (+23%)

Case A) Case B)

Drag force Scour dept Drag force Scour dept

ACKNOWLEDGMENTS This research was possible with the support of Regione Lombardia - Dote Ricercatori fellowship assigned to one of the authors. REFERENCES Amaya, A.J.R., Lengerke, O., Cosenza, C.A., Dutra, M.S. and Tavera, M.J.M. (2009). Comparison of Defuzzification Methods: Automatic Control of Temperature and Flow in Heat Exchanger, Automation Control - Theory and Practice, Edited by: A D Rodi. Apelt, C.J. and Isaacs, L.T. (1968). Bridge piers - hydrodynamic force coefficients, Journal of Hydraulics Div., Vol. 94 (HY1), 17-30. Ballio, F., Bianchi, A., Pranzetti, S., De Falco, F. and Mancini M. (1999) (in italian). Vulnerabilit idraulica dei ponti fluvial, Atti del XXVI Convegno di Idraulica e Costruzioni Idrauliche, Catania, Vol. 3, 69-79. Biondini, F., Bontempi, F. and Malerba, P.G. (2004). Fuzzy Reliability of Concrete Structures, Computers & Structures, Vol. 4(82), 1033-1052. Blevins, R.D. 1984. Applied Fluid Dynamics handbook, New York: Van Nostrand Reinhold Co. Da Deppo, L. and Datei, C. (1994) (in Italian). Le opere idrauliche nelle costruzioni stradali, Cosenza: Editoriale Bios. Da Deppo, L., Marson, C., Fioretto, V. and Caroni, E. (2006). Clear water scour analysis around groynes and bridge abutments, Proceedings of the Third Edition of the International Conference on Fluvial Hydraulics. Dordoni, S., Malerba, P.G., Sgambi, L. and Manenti S. (2010). Fuzzy reliability assessment of bridge piers in presence of scouring. Proceedings of The Fifth International Conference on Bridge Maintenance, Safety and Management (IABMAS10), Philadelphia (USA), 11-15 July 2008. 2010 Taylor & Francis Group, London. Elishakoff, I. (1983). Probabilistic theory of structures. Published by John Wiley and Sons. Elith, J., Burgman, M.A. and Regan, H.M. (2002). Mapping epistemic uncertainties and vague concepts in predictions of species distribution, Ecological Modelling,Vol. 157, 313-329. FHA, U.S. Dept. of Transp. (2001). Evaluating Scour At Bridges, Fourth Edition, Publication No. FHWA NHI 01-001, May 2001. Goode, J.K. and Burbank, D.W. (2009). Numerical study of degradation of fluvial hanging valleys due to climate change, Journal of Geophysical Research, Vol. 9(114).

Hamill, L. (1999). Bridge Hydraulics. New York: E&FN Spon. Holicky M. (2010). Fuzzy probabilistic models for structural serviceability, Journal of Applied Quantitative Methods, Vol. 5(3), 467-478. Kala, Z. (2007), Fuzzy Probabilistic Analysis of Steel Structure Focused on Reliability Design Concept of Eurocodes, Nonlinear Analysis: Modelling and Control, Vol. 12(3), 371382. Kala, Z. (2007). Stability problems of steel structures in the presence of stochastic and fuzzy uncertainty, Thin-Walled Structures, Vol. 45, 861865. Kwakernaak, H., (1978). Fuzzy random variables-i. denitions and theorems, Information Sciences, Vol. 15, 129. Kwakernaak, H., (1979). Fuzzy random variables-ii. algorithms and examples for the discrete case, Information Sciences, Vol. 17, 253278. Laursen, E.M. (1962). Scour at bridge crossings, Transaction of the American Society of Civil Engineers, Vol. 62(127), 166-180. Li, X. and Liu, B. (2006). New independence denition of fuzzy random variable and random fuzzy variable, World Journal of Modelling and Simulation, Vol. 2(5), 338-342. Miglio, A., Gaudio, R. and Calmino, F. (2009). Mobile-bed aggradation and degradation in a narrow flume: Laboratory experiments and numerical simulations, Journal of Hydroenvironment Research, Vol. 9(3), 9-19. Mller, B., Graf, W., Beer, M. and Sickert, J.U. (2002). Fuzzy Randomness - Towards a new Modeling of Uncertainty, Fifth World Congress on Computational Mechanics July 712, 2002, Vienna, Austria. Muzzammil, M., Siddiqui, N.A. and Siddiqui, A.F. (2008). Reliability considerations in bridge pier scouring, Structural Engineering & Mechanics, Vol. 8(28). Piegat A. (2001). Fuzzy Modeling and Control, Physica-Verlag Heidelberg, New York. Puri, M.L. and Ralescu D. (1986). Fuzzy random variables. Journal of Mathematical Analysis and Applications, Vol. 114, 409422. Regan, H.M., Colyvan, M. and Burgman, M.A. (2002). A taxonomy and treatment of uncertainty for ecology and conservation biology, Ecological Applications, Vol. 12(2), pp. 618628. Richardson, E.V. and Davis, S.R. (2001). Evaluating Scour at Bridges, Hydraulic Engineering Circular No. 18, US Department of Transportation. Sgambi, L. (2004). Fuzzy approach in the three-dimensional non linear analysis of reinforced concrete two-blade bridge piers, Computers & Structures, Vol. 4(82), 1067-1076. Sobhani, J. and Ramezanianpour A.A. (2009). Uncertainty analysis of corroded reinforced concrete structure, Asian Journal of Civil Engineering (Building and Housing), Vol. 10(2), 145-161. Zadeh, L.A. (1965). Fuzzy algorithms, Information and Control, Vol. 65(12), 94-102. Zadeh, L.A. (2002) Foreword of T. Ross, J. M. Booker and W. J. Parkinson (Eds), Fuzzy Logic and Probability Applications: Bridging the Gap, ASA SIAM Zadeh, L.A., (1978). Fuzzy sets as a basis for a theory of possibility, Fuzzy Sets and Systems, Vol. 78(1), 3-28.

Vous aimerez peut-être aussi