Vous êtes sur la page 1sur 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/274295037

Effect of Deflection Increment Size on Apparent Performance of Fiber


Reinforced Concrete Beams and Panels

Article  in  Journal of ASTM International · July 2011


DOI: 10.1520/JAI103248

CITATIONS READS

0 181

3 authors, including:

Erik Bernard
BG&E
43 PUBLICATIONS   476 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Erik Bernard on 26 September 2016.

The user has requested enhancement of the downloaded file.


Journal of ASTM International, Vol. 8, No. 7
Paper ID JAI103248
Available online at www.astm.org

E. S. Bernard1 and R. Coleman2

Effect of Deflection Increment Size on Apparent Performance


of Fiber Reinforced Concrete Beams and Panels

ABSTRACT: Post-crack performance testing of Fiber Reinforced Concrete (FRC) is commonly undertaken
using beam tests, such as ASTM C1609/C1609M, or panel tests, such as ASTM C1550. Both these tests
demand the use of sophisticated servo-controlled test machines with data acquisition systems that record
several hundreds to thousands of points in a typical test. Existing standard test methods include detailed
specifications regarding how load should be applied and how deformation is measured, but they retain sim-
ple prescriptions regarding data acquisition and the total number of points required to adequately character-
ize the performance of a specimen. The present investigation has been undertaken to assess the effect of
changes in the size of the deflection increment between recorded points in a load-deflection curve on the
apparent outcome of a test. The results indicate that the maximum increment of deflection between succes-
sively acquired data points should be 0.020 mm for a ASTM C1550 panel test and 0.0035 mm for a ASTM
C1609/C1609M beam test in order to limit potential errors in energy absorption and peak load to within ac-
ceptable levels. This is smaller than the current recommendation of 0.05 mm for ASTM C1550 and thus a
change to a smaller measurement interval is recommended. For ASTM C1609/C1609M the currently rec-
ommendation for rate of data acquisition over the first span/900 deflection is adequate but thereafter the
rate of acquisition should be increased by 50%.
KEYWORDS: fiber reinforced concrete, post-crack performance, testing, data acquisition

Introduction
Post-crack performance assessment tests for fiber reinforced concrete (FRC) and shotcrete (FRS) have
evolved over many years to the point where ASTM C1609/C1609M [1] and ASTM C1550 [2] now require
sophisticated servo-controlled test machines to adequately control unstable behavior immediately after crack-
ing of the concrete matrix. These machines are capable of high response rates based on complex electronic
feedback systems that also permit many variations in the method of performance measurement. Given that
numerous manufacturers offer test machines suitable for post-crack performance assessment (also known as
“toughness” testing) of FRC, and test machine technology has evolved extensively over several decades,
there exist many variations on the exact means of control and data acquisition. Unlike other standard tests
performed on concrete, such as C39/C39M [3] for compressive strength and C78 [4] for flexural strength,
both of which generate just one estimate of load resistance (the peak load resistance, taken to represent the
strength of the concrete), toughness tests generate a load-deflection record that will include several hundreds
to thousands of data points. Not only is each data point subject to sampling errors arising from limitations in
measurement precision, but the density of data recording (i.e., the sampling interval) has associated with it
intrinsic errors that influence the validity of the record’s representation of the continuous response of the
specimen. An assessment of the effect of changes in the displacement increment size between data points on
the apparent performance of a specimen is of value in understanding the sources of error that contribute to
the high variability typically observed in toughness testing, especially for beam tests [5,6].

Background
Numerous toughness tests exist for the assessment of post-crack performance of FRC [7]. The present
investigation has been restricted to consideration of testing using ASTM C1609/C1609M and ASTM
C1550 because these are broadly representative of most available beam and panel-based tests.

Manuscript received July 12, 2010; accepted for publication May 5, 2011; published online July 2011.
1
TSE Pty. Ltd., P.O. Box 763, Penrith, NSW 2751, Australia.
2
Institute for Marine and Antarctic Studies, Univ. of Tasmania, Private Bag 129, Hobart TAS 7001, Australia.

Copyright V
C 2011 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.
2 JOURNAL OF ASTM INTERNATIONAL

FIG. 1—ASTM C1609/C1609 M third-point loaded beam test.

ASTM C1609/C1609M involves a simply-supported third-point loaded beam tested in such a way that
the central deflection of the beam increases at a constant rate of between 0.10 and 0.30 mm/min depending
on the level of deformation attained (Fig. 1). This test requires closed-loop control of the loading mecha-
nism based on feedback from linear voltage displacement transducers (LVDTs) used to measure the cen-
tral deflection of the specimen relative to the supports. If load is applied using an hydraulic piston,
delivery of the hydraulic fluid to the piston must be undertaken via an electronically controlled servo-valve
because other methods of closed-loop fluid control, such as proportional valves, are too slow to adequately
control instability associated with rapid crack propagation through the concrete matrix [6].
ASTM C1550 involves the application of a central point load to a round panel resting on three sym-
metrically arranged pivoted points (Fig. 2). Load is imposed at a constant rate of displacement using a
servo-controlled test machine. However, unlike ASTM C1609/C1609M, the user has the choice of apply-
ing load based on closed-loop servo-control [8] of either the central deflection of the panel or displacement
of the actuator. Both mechanisms demand measurement of displacement, either that of the actuator or the
specimen, to use as electronic feedback for the control loop.
The requirement for an electronically controlled servo-mechanism in both ASTM C1609/C1609M
and ASTM C1550 demands advanced electronics capable of rapid measurement of specimen or actuator
displacement. The software associated with such electronics typically permits data acquisition based on
specified increments either of time, load, actuator displacement, or a strain (such as the measured central
deflection of the specimen). Unfortunately, there exists little consistency between standards regarding the
required data interval. In the 2010 version of ASTM C1609/C1609M, the control parameter for execution
of the test is the measured deflection of the center of the beam, but the criterion determining the data ac-
quisition rate is time (a minimum acquisition rate of 1 Hz or 2.5 Hz is required, depending on the deflec-
tion). In the 2010 version of ASTM C1550, the control parameter for execution of the test is either the
measured deflection of the actuator or the central deflection of the specimen, but the criterion determining
data acquisition is not specified. Instead, the load-deflection record is required to comprise at least 800
points between the start of loading and a central deflection (or actuator deflection) of 40 mm. This lack of
specificity in the required data interval can lead to variations in the apparent performance of the specimen
as will be demonstrated below.
BERNARD AND COLEMAN ON EFFECT OF DEFLECTION INCREMENT SIZE 3

FIG. 2—ASTM C1550 round panel test.


Sources of Error
Each data point acquired during a single test is subject to several types of error arising from limitations in
the precision of the original measured parameter, analogue to digital (A/D) conversion errors, and round-
ing errors within the data record [9]. Time is usually the most accurately measured parameter but this pa-
rameter is seldom recorded with its original (raw) level of precision which is in the order of nano-seconds.
Typically, rounding (or truncation) errors can arise in this parameter which affect the apparent rate of
loading or deflection because this truncated time value is then used in subsequent calculations.
In the discussion that follows, only errors associated with the testing of individual specimens are
addressed. This study has not been concerned with errors that may arise from variations in material proper-
ties between a number of nominally identical specimens.
Most modern test machines measure test parameters, such as load or displacement, as an analogue
electrical signal that is converted to digital data via a high speed analogue-to-digital converter. The ana-
logue output from the measurement device (load cell or displacement transducer) includes errors due to
limitations in machine calibration precision and electrical noise in the signal. The precision of the A/D
converter depends on the number of bits used on the digital side of the system (24 bit in the present inves-
tigation), but even high bit systems introduce an error associated with discretization. Both these types of
error (A/D conversion and rounding) manifest themselves in every data point recorded during a test. The
rate of data conversion is usually related to the feedback loop speed which typically lies in the range 500–
6000 Hz (anecdotal evidence indicates that rates slower than this are insufficient for effective control of
unstable post-crack behavior). Data are continuously produced at the peak rate and compared to the
selected acquisition criterion (for example, the minimum change in time or displacement since the last
reading) to determine whether each point is recorded or ignored. Once data have been acquired, it is usu-
ally transferred to a spreadsheet in which it is manipulated to find the load resistance at specified deflec-
tions, the peak load resistance, and other derived parameters such as residual strength, equivalent residual
strength, or relative toughness [10].
It must be noted that the performance of the specimen is assessed in terms of load and deflection, not
time, and thus errors in performance will be directly related to errors in load and deflection measurements
and only indirectly related to errors in other measured parameters such as time. That is why it is preferable
to specify the output from a test machine in terms of the maximum acceptable increment of deflection
rather than rate of data acquisition based on time. How this limit on the maximum acceptable deflection
increment size will be effected for a particular machine will depend on the operating system used since
the manner in which data are acquired varies significantly between machines (for example, some will only
permit a time-based rate of acquisition to be specified, while others will permit the size of displacement in-
crement to be specified without reference to time).
Energy absorption is normally found as the integral under the load-deflection curve which is calcu-
lated as the sum of products of successive increments of deflection and the associated load resistance.
4 JOURNAL OF ASTM INTERNATIONAL

FIG. 3—Curtailment of energy absorption at measurement end points for a measurement interval d repre-
sented by n þ1 points.

Thus, the area under the curve is effectively found as the sum of numerous slender trapezoids. However,
load resistance is seldom measured at precisely the required end point deflections and thus some discrep-
ancy will occur between the desired and actual deflections leading to partial increments both at the initial
and final measured points (Fig. 3). Most spreadsheets curtail (that is, ignore) these partial areas and thus
the calculated energy under the curve is always slightly less than the “true” energy that would be calcu-
lated if the end points were included. As the average increment size increases in magnitude relative to the
total deflection measurement interval, the error associated with this curtailment becomes more significant.
A second source of error associated with increment size is related to the “roughness” (and shape) of
the load-deflection profile. When a continuous load-deflection record is represented by discrete points, the
true area under the curve is approximated by a series of trapezoidal areas whose sum diverges from the
true area in a manner that depends on the linear versus non-linear shape of the profile (Fig. 4). As the num-
ber of points approximating the record falls, the relative magnitude of the error usually increases, although
this may not necessarily occur for particularly smooth or flat curves.
While the influence of the magnitude of the displacement increment between data points on apparent
errors identified above can be assessed theoretically, experimental confirmation of the change in perform-
ance is of value so that the magnitude of errors due to discretization and curtailment that are typical of real
beam and panel tests can be assessed. This has been sought as outlined below.

FIG. 4—Errors introduced through discretization of a continuous record of length d in n intervals.


BERNARD AND COLEMAN ON EFFECT OF DEFLECTION INCREMENT SIZE 5

Experimental Program
The investigation involved testing 12 ASTM C1609/C1609M beams and 18 ASTM C1550 round panels
to generate post-crack performance data that could be assessed to determine the effect of data acquisition
rate on apparent performance. The specimens were taken from commercial Quality Assurance testing pro-
grams and represented both steel and macro-synthetic FRC. The performance of the ASTM C1609/
C1609M beams spanned a range from 59.03 to 104.04 Js energy absorption at 3 mm deflection. The per-
formance of the ASTM C1550 panels spanned a range from 46.8 to 138 Js at 5 mm central deflection, and
322 to 609 Js at 40 mm central deflection. This span represents the typical range of performances for FRC
in commercial use within Australia. An MTS Flextest GT controller operating at 6000 Hz was used to-
gether with an MTS 244.31 250 kN hydraulic actuator for the beam tests or an MTS 244.21 100 kN hy-
draulic actuator for the panel tests.
The tests were performed in conformance with standard practice except that the rate of data acquisi-
tion was much greater than normal (by a factor of 5). The panel tests were performed with displacement of
the actuator used as the criterion for data acquisition, and since this parameter was also used to control
execution of the test, each record resulted in the same number of data points (about 4500). The beam tests
were performed with central deflection (as measured by one of the two side-mounted LVDTs) used as the
controlling parameter, but displacement of the actuator was used as the criterion for data acquisition. This
is equivalent to using time as the acquisition parameter since the actuator moved at a more-or-less constant
rate. Since the relation between displacement of the actuator and the central deflection of the beam is influ-
enced by the load train stiffness and the location of the crack (which varies between specimens), the total
number of data points for each beam ranged between about 5000 and 7500.
Only the post-crack portion of the load-deflection record for each specimen has been assessed in this
investigation because the focus of error assessment has been post-crack energy absorption. Preliminary
work also indicated that the rate of displacement could change immediately after the concrete matrix
cracks. The prescribed rate of displacement for the beams was also deliberately varied between 0.10 and
0.25 mm/min over the first 0.25 mm of each test [1]. Since these deliberate changes interfered with the
size of the increment between data points in the early stages of the tests, thereby making it difficult to
assess the effect of a uniform acquisition process, the initial part of each load-deflection record has been
ignored.

Data Processing
Once each test had been completed the data were used to generate estimates of performance for each sin-
gle specimen. To examine the effect of a change in deflection increment size between each acquired point
on apparent performance, each of the 30 load-deflection records was successively decimated so that, for
example, every second point throughout the record was retained, then every third point, fourth point, etc.
through to every thousandth point. The result was about 15 decimated subsets of load-deflection data for
each of the 30 original data records. This decimation was equivalent to changing the magnitude of the
deflection increment between successively recorded points in the original tests because the value of each
and every point was not influenced by the selection criterion governing whether they were retained or not.
Once these subsets had been obtained they were copied into a calculation spreadsheet and used to find the
peak load and energy absorption.

Results
Each test resulted in a load-deflection curve of the type shown in Fig. 5 for ASTM C1609/C1609M beams
and ASTM C1550 panels. These figures also show the rate of deflection between successive data points
plotted as crosses and referenced to the right hand axis. This rate of deflection was simply calculated as
the difference in deflection between successive points divided by the elapsed time interval. The rate of
deflection varied substantially for each beam, especially when small loops occurred in the load-displace-
ment curve. These loops occurred in response to crushing of detritus on the surface of the beam or sudden
falls in load resistance due to slipping or rupture of fibers bridging the crack. Elastic relaxation of the load
train and the uncracked portions of the beam caused these sudden falls in load resistance to generate rever-
sals in displacement as the feedback loop attempted to compensate for the change in resistance. Some
6 JOURNAL OF ASTM INTERNATIONAL

FIG. 5—Load resistance and rate of displacement as a function of central deflection for a typical (a)
C1609 beam test, and (b) C1550 panel test.

beams exhibited more loops than others, and the magnitude of the loops varied substantially even for the
same beam; hence a variation arose in average rate of displacement between specimens. This clearly vio-
lated the requirement to maintain a nominally uniform rate of loading as specified in the standard, but
there is little that can be done to prevent this other than maintain tight control over tuning of the servo-
loop.
Inspection of Figs. 6 and 7 reveals several aspects of the influence that changes in data acquisition had
on the apparent post-crack performance of the beams and panels. Firstly, the apparent maximum load fell
as the average deflection increment size increased (Fig. 6). This occurred because an increase in deflection
increment between successively recorded points increased the chance that the “true” maximum load resist-
ance was missed [11]. The significance of missing a maximum in the load record depended on whether the
maximum was associated with a pronounced peak in load resistance or a plateau: obviously, greater errors
occurred if the maximum was associated with a sharp peak in the curve. Apparent energy absorption fell
due to the curtailment phenomenon as the number of points comprising the record was reduced with sub-
stantial discretization errors arising when the total number of data points fell below about 50 (Fig. 7). In

FIG. 6—Influence of deflection increment size between successive data points on relative max load for (a)
C1609/C1609M beam tests, and (b) C1550 panel tests.
BERNARD AND COLEMAN ON EFFECT OF DEFLECTION INCREMENT SIZE 7

the current version of ASTM C1550 the smallest permissible number of points in the measurement interval
up to 5 mm central deflection is 100 points (based on the prescribed minimum of 800 points over the full
40 mm and an assumed uniform rate of data acquisition throughout a test). Thus the error in energy
absorption will be about 2% due solely to the curtailment phenomenon.

Discussion
The present results indicate that apparent energy absorption is influenced by the number of points compris-
ing the data series (Fig. 7). The relative magnitude of the effect is similar for both the beams and panels
and appears to be due primarily to curtailment of energy summation at the end points of the load-deflec-
tion record (a result of algorithms commonly used within spreadsheets and other programs) together with
discretization errors that are difficult to avoid when representing a continuous record by discrete points.
To examine this phenomenon for a general load-displacement record, the relationship between the number
of data points (defined as n þ1 in the measurement interval) and the resulting error can be assessed as
follows.
The effect of curtailment at the endpoints on error in energy absorption can be determined with refer-
ence to Fig. 3. Ignoring, for the time being, the irregular nature of a typical curve between the end points
(i.e., assume a constant slope between the end points), the total area under a curve between the end points
is equal to
d
A ¼ Pða þ 1Þ (1)
2

where d is the total displacement over which energy is summed, P is the measured load at the end point of
greater load resistance, and a is a reduction factor accounting for the average slope of the curve. Assuming
the loads at the curtailment points are the same as at the end points, the area remaining after curtailment A
is found as
 
n1
A  Pða þ 1Þd (2)
2n

hence the relative area remaining after curtailment is equal to

A n  1
¼ (3)
A n

FIG. 7—Influence of number of data points within the measurement interval on relative energy absorption
for (a) C1609/C1609M beam tests, and (b) C1550 panel tests.
8 JOURNAL OF ASTM INTERNATIONAL

This simple expression is based on the assumption that the load at the end points is the same as the load at
the curtailment points. This equation breaks down if the curve is non-linear and hence is an approximate
estimate that is only applicable for small displacement increments where a linear slope between points
approximates the curve.
The effect on the error in apparent energy absorption of discretizing a continuous curve into a finite
number of intervals can be assessed with reference to Fig. 4. The area under a continuous function, repre-
sented by y ¼ P(x), can be written as
ðb
A ¼ PðxÞdx (4)
a

where a, b represent the limits of integration. If the continuous function is known, then an analytical
expression for the area can be found provided the equation is integrable. This is not the case for situations
where we have discrete data. In this instance, techniques such as the composite Simpson’s or Trapezoidal
Rule are required [12], and approximate formulae for the errors in these two rules have been given by
Cruz-Uribe and Neugebauer [13]. If, over an interval of integration a ) b, where d ¼ b  a, we sub-divide
this interval into n smaller sub-intervals, then the composite trapezoidal rule can be written as
ðb
ba
A ¼ PðxÞdx  ðPðx0 Þ þ 2Pðx1 Þ þ 2Pðx2 Þ þ …2Pðxn1 Þ þ Pðxn ÞÞ (5)
2n
a

If the number of data points in the displacement record is decimated (that is, reduced) from n þ 1 to l þ 1,
and the number of intervals from n to m, the area A0 of this new discretized displacement record would be
d
A0 ¼ ðPðx0 Þ þ 2RPðxl Þ þ Pðxn ÞÞ (6)
2m

where the subscript l accounts for every lth point in m intervals over d. Note that n and m have an even
number of intervals. The difference in area (AA0 ) is equal to
dðm  nÞ d
A  A0 ¼ ðPðx0 Þ þ 2RPðxcommon Þ þ Pðxn ÞÞ þ ðRPðxmissing ÞÞ (7)
2mn n

where the subscript “common” refers to points that are included in both the full and discretized series, and
“missing” refers to points that occur in the full series but are missing from the discretized series. Follow-
ing some rearrangement, the relative area under the curve following discretization is therefore
A nð1  cÞ
¼ (8)
A m

where
P
2 Pðxmissing Þ
c¼ P P (9)
Pðx0 Þ þ Pðxn Þ þ 2 Pðxcommon Þ þ 2 Pðxmissing Þ

The parameter c is related to the shape of the original curve. Given that the error in energy absorption
arises as a result of both curtailment and discretization, each of which contribute varying proportions to
the total error depending on the shape of the curve and “roughness” of the data, the total error  can be
approximated as being proportional to
n0  b
¼ (10)
n0

in which n’ is the number of points in the final discretized series. Curves of this form are shown as the
“approximating function” in Fig. 7 and reveal that b 1.3 for beams and 2.3 for panels. The magnitude of
BERNARD AND COLEMAN ON EFFECT OF DEFLECTION INCREMENT SIZE 9

b probably changes with each data set used as it is dependent on the characteristics of each load-deflection
curve. The values of b listed above have therefore been obtained by curve-fitting Eq. (10) to the available
data for each test method to get the most robust estimates.
The error in energy absorption due to a large increment size between data points is particularly notable
for small overall measurement intervals (i.e., the interval between the beginning and end of performance
assessment). This is apparent on examination of the data shown in Fig. 8 which shows the effect of deci-
mating the data record for energy absorption obtained from the ASTM C1550 panel tests. The smallest
commonly specified central deflection over which energy absorption is summed for the ASTM C-1550
panel test is 5.0 mm and the largest is 40 mm. Figure 8(a) shows that the energy absorption at 5 mm falls
more sharply when the number of points in the load-deflection record is reduced by a factor of 100 than
happens for energy absorption at larger central deflections. This is because the number of points compris-
ing the final record is smaller for the 5 mm energy absorption estimate. Each of the 5 mm energy estimates
were based on an average of 526 original sampled points, each 10 mm energy estimate was based on about
1055 sampled points, each 20 mm energy estimate was based on about 2110 sampled points, and each 40
mm energy estimate was based on 4221 sampled points. After decimation by a factor of 100, the number
of points upon which the energy estimates were based was reduced to 5, 10, 21, and 42, respectively.
Hence the change in energy was much more pronounced for the 5 mm estimate than for the 40 mm esti-
mate because the sampling error had moved to the far left of the curve shown in Fig. 7(b). The change in
energy had nothing to do with material properties but was instead due primarily to curtailment at both
ends of the data record.
Figure 8(b) shows that when the number of points in each data record (for energy absorption at 5, 10,
20, and 40 mm central deflection) is the same, a consistent reduction in the number of points (in this case,
from about 520 to about 11 points) results in a similar fall in apparent energy absorption. This corroborates
the trend shown in Fig. 8(a). Both these results are entirely consistent with the trend shown in Fig. 7(b). A
similar phenomenon occurred for energy absorption in the ASTM C1609/C1609M beams, but the range of
performances displayed by the available beams was not as broad as for the panels and thus the trend was
not so apparent.
It is also evident in Fig. 8(b) that the change in energy absorption as the number of data points is var-
ied appeared independent of the actual performance of the specimen (at least for energy absorption in the
range presently investigated). This indicates that errors due to discretization and curtailment are not
related to the performance of the specimens being assessed. This is because discretization and curtailment
are purely numerical procedures that are invariant to the origin and nature of the numerical data being
processed. The lack of a relation between errors due to discretization and curtailment and the performance

FIG. 8—Influence of magnitude of energy absorption on the fall in energy absorption for C1550 panels as
(a) data are decimated by a factor of 100, and (b) number of data points is reduced from about 520 to
about 11 points.
10 JOURNAL OF ASTM INTERNATIONAL

of the FRC specimens is consistent with the general observation that the magnitude of errors in multi-pa-
rameter processes are not related in a predictable manner to the magnitude of the parameters that are sub-
ject to error [14].
The minimum number of points required to avoid curtailment errors in either the ASTM C1609/
C1609M beam test or the ASTM C1550 panel test should be based on the smallest measurement intervals
required and not the maximum deflections sustained in the tests. If users specify performance at even
smaller displacement intervals (that is, less than 0.5 mm in ASTM C1609/C1609M or 5.0 mm in ASTM
C1550) consideration must be made to the effect on the accuracy of energy absorption. If a 1% error in
energy absorption is taken to be the maximum acceptable error due to discretization and curtailment (see
Fig. 7), then Eq. 10 indicates that a minimum of about 130 points must be used between end points across
the smallest measurement interval to ensure errors in energy absorption in ASTM C1609/C1609M beam
tests due to this phenomenon remain acceptably small. This corresponds to a maximum allowable 0.0035
mm deflection increment for ASTM C1609/C1609M. The current version of ASTM C1609/C1609M
specifies a time-based rate of data acquisition of at least 2.5 Hz, but this can be reduced to 1 Hz after a
deflection of span/900 is exceeded. The rate of deflection is allowed to vary between 0.10 mm/min before
span/900 is reached and 0.30 mm/min thereafter (corresponding to a deflection interval size of between
0.00067 and 0.005 mm). This indicates that the specified 2.5 Hz rate of data acquisition is quite adequate
for the initial part of the load-deflection curve up to span/900, but the rate of 1 Hz thereafter is too slow
and should be increased to at least 1.5 Hz (or a maximum 0.0035 mm deflection between points).
The approximating function Eq. 10 also indicates that a minimum of about 230 points must be used in
the smallest measurement interval for the ASTM C1550 panel test to ensure the error due to curtailment
and discretization is less than 1%, suggesting a maximum allowable 0.020 mm deflection increment for
ASTM C1550 over the first 5 mm of deformation. This corresponds to a sampling rate of 3.33 Hz, which
is close enough to 4 Hz if fractional sampling rates are not available for a particular data acquisition sys-
tem. The current version of ASTM C1550 requires at least 100 points per 5 mm (assuming uniform point
spacing throughout) which corresponds to 0.05 mm deflection between points. This appears to be inad-
equate, so the minimum number of uniformly spaced points in the load-deflection record should be
increased to 250 for the first 5 mm, or 2000 for the entire 40 mm interval if only one interval size can be
specified for the whole test.
The recommended changes in maximum measurement interval between samples (or minimum sam-
pling frequency) are easy to implement on most test machines. If rates of 1.5 Hz for ASTM C1609/
C1609M or 3.33 Hz for ASTM C1550 cannot be achieved, then a slightly smaller measurement interval or
faster acquisition rate should be implemented. These changes will result in a larger number of points
acquired for each test but the increase is within manageable limits for most operating systems.

Conclusions
This investigation has revealed that a reduction in displacement increment size between recorded data
points results in improved estimates of maximum load and energy absorption for ASTM C1609/C1609M
beams and ASTM C1550 panels. However, a smaller increment size also leads to increased data storage
requirements. Based on results obtained for a variety of steel and macro-synthetic Fiber Reinforced Con-
crete ASTM C1609/C1609M beams and ASTM C1550 panels, it is recommended that the maximum size
of deflection increment between acquired data points be 0.020 mm for ASTM C1550 panel tests and
0.0035 mm for ASTM C1609/C1609M beam tests to maintain errors in peak load and energy absorption
due to discretization and curtailment of less than 1%. This means that the minimum number of acquired
points in an ASTM C1550 test (if uniformly spaced) should be 2000 (compared to the present requirement
for 800). For ASTM C1609/C1609M, the minimum rate of data acquisition should be increased to 1.5 Hz
from the present 1 Hz. Both these changes are easy to implement.

Acknowledgments
The writers wish to thank Josh Tolhurst for undertaking the tests and Geoff Xu for his assistance in deci-
mating the resulting load-deflection curves. Additional guidance on machine operation was provided by
the MTS Corporation, Eden Prairie, MN.
BERNARD AND COLEMAN ON EFFECT OF DEFLECTION INCREMENT SIZE 11

References

[1] ASTM C1609/C1609M, “Standard Test Method for Flexural Performance of Fiber Reinforced Con-
crete (Using Beam with Third-Point Loading),” Annual Book of ASTM Standards, Vol. 04.02, ASTM
International, West Conshohocken, PA, pp. 846–854.
[2] ASTM C1550, “Standard Test Method for Flexural Toughness of Fiber Reinforced Concrete (Using
Centrally Loaded Round Panel),” Annual Book of ASTM Standards, Vol. 04.02, ASTM International,
West Conshohocken, PA, pp. 768–781.
[3] ASTM C39/C39M, “Standard Test Method for Compressive Strength of Cylindrical Concrete Spec-
imens,” Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA.
[4] ASTM C78, “Standard Test Method for Flexural Strength of Concrete (Using Simple Beam with
Third-point Loading),” Annual Book of ASTM Standards, ASTM International, West Conshohocken,
PA.
[5] Bernard, E. S., “Correlations in the behavior of fibre reinforced shotcrete beam and panel spec-
imens,” Mater. Struct., Vol. 35, 2002, pp. 156–164.
[6] Bernard, E. S., “Influence of Test Machine Control Method on Flexural Performance of Fiber Rein-
forced Concrete Beams,” J. ASTM Int., Vol. 6, No. 9, 2009.
[7] Mindess, S., Taerwe, L., Lin, Y.-Z., Ansari, F., and Batson, G., “Standard Testing,” Chapter 10, High
Performance Fiber Reinforced Cement Composites 2, A. E. Naaman and H. W. Reinhart, Eds.,
E&FN Spon, London, 1995, pp. 384–421.
[8] Anderson, W. R., Controlling Electrohydraulic Systems, Marcel Dekker, NY, 1988.
[9] Himelblau, H., PierSol, A. G., Wise, J. H., and Grundvig, M. R., Handbook for Dynamic Data Acqui-
sition and Analysis, IEST, Rolling Meadows, IL, 2006.
[10] Tatnall, P. C., “Significance of Tests and Properties of Concrete and Concrete-Making Materials,”
Chapter 49, STP 169D, J. F.Lamond and J. H. Pielert, Eds., ASTM International, West Consho-
hocken, PA, 2006.
[11] Wise, J. H., “The Effects of Digitizing Rate and Phase Distortion Errors on the Shock Response
Spectrum,” Proc. IES, 1983, pp. 36–43.
[12] Atkinson, K. A., An Introduction to Numerical Analysis, 2nd ed., John Wiley and Sons, New York,
1989.
[13] Cruz-Uribe, D. and Neugebauer, C. J., “Sharp error bounds for the trapezoidal rule and Simpson’s
rule,” J. Inequal. Pure Appl. Math., Vol. 3, Issue 4, 2002, pp. 1–22.
[14] Kirkup, L. and Frenkel, R. B., An Introduction to Uncertainty in Measurement, Cambridge Univ.
Press, New York, 2006.

View publication stats

Vous aimerez peut-être aussi