Professional Documents
Culture Documents
FraMCoS-X
G. Pijaudier-Cabot, P. Grassl and C. La Borderie (Eds)
https://doi.org/10.21012/FC10.233352
Abstract. Experimental results of fatigue tests in concrete present a wide scatter, with a difference
of even several orders of magnitude between the highest and lowest obtained fatigue lives. The
behaviour of fatigue results can be described by a probability model like the Weibull distribution.
However, it is not clear how many tests are necessary to properly estimate the distribution parameters
for a given concrete and loading conditions. In this work, a method has been developed so that the
number of tests can be selected on the basis of the admissible error for the characterisation of the
probabilistic behaviour of fatigue. As support for the performed study, a fibre-reinforced concrete
was produced and 105 cubic specimens were tested. Afterwards, the sampling method known as
‘bootstrap’ was applied, which consists in generating multiple random samples of a given number
of fatigue life results taken from the experimental set. The difference between the distribution fitted
to each sample and the reference distribution of the complete set of values is the error made by the
sample to characterise the material. The distribution followed by the error itself after all the sample
simulations allows to determine the expected maximum error for a given safety level in relation to the
number of tests. Different sample sizes from 3 to 90 tests have been analysed. The results provided
by the method also allow to define a fatigue design distribution for the studied concrete and loading
conditions.
1
José J. Ortega, Gonzalo Ruiz, Rena C. Yu, Nelson Afanador-Garcı́a, Manuel Tarifa, Elisa Poveda and Xiaoxin Zhang
tributions, the Weibull function represents very how many tests must be performed in order to
well the mechanical fatigue of concrete [1–3]. have an error under an admissible value, both as
However, there are no general criteria for se- a maximum error or at a given failure probabil-
lecting the minimum number of tests that must ity. Moreover, a fatigue design distribution can
be performed in order to properly reproduce be defined so that the fatigue resistance is not
the behaviour of concrete under fatigue and to overestimated for the specified safety level.
know the upper and lower limits of the fatigue In the next section the material and the ex-
life range. That large scatter of results implies perimental programme are described; in Sec-
that an elevated number of tests could be nec- tion 3, the followed method for the statistical
essary. Due to the cost in time and resources, analysis is explained; Section 4 shows the ob-
fatigue characterisation usually keeps to a re- tained results and, finally, the conclusions are
duced number of tests or to the deterministic presented in Section 5.
expressions given by design standards [4, 5],
which provide a mean value of cycles for the 2 MATERIAL AND TESTS
specified stress level. A steel-fibre-reinforced self-compacting
The probability distribution obtained from a concrete was produced. The cement type was
small sample of specimens can greatly differ CEM I 52.5 SR and the aggregates were sili-
from the distribution to which the entire mate- cious with a maximum size of 8 mm. The fibres
rial tends. That difference is the error made in were straight of 13 mm in length and 0.20 mm
the characterisation, which could reach an inad- in diameter and its proportion in the mix was
missible value without the awareness of being 0.2% in volume.
in that situation. The concrete was poured in prismatic
Some authors have proposed different meth- moulds of 400 × 100 × 100 mm3 . The spec-
ods to define the necessary sample size [6–9]. imens for fatigue testing were cut from those
However, distributions other than Weibull are prisms as cubes of 40 mm side.
used in some cases and, in general, the meth- Tests were performed in a servo-hydraulic
ods are relatively complex and use coefficients machine with a maximum capacity of 250 kN.
given for particular probability levels. The machine has two loading platens between
In the present work [10], it is intended to which specimens are disposed and a ball-and-
show in a simpler manner the relationship be- socket joint to align the load with the machine
tween the number of performed tests and the axis. This joint was fixed in each test once the
maximum committed error. Results are given platens were accommodated to the specimen
for different safety levels and along the range of faces at a pre-load of 4 kN. Besides, an addi-
failure probability. The study starts from a set tional self-produced ball-and-socket joint with
of 100 fatigue tests carried out with specimens a base of the same square dimensions that the
made with the same concrete and with the same specimen faces was placed on each one. With
loading configuration. By means of statistical this device, the load is centred on the concrete
sampling techniques, the distribution and range cube, reducing the effect of possible eccentrici-
of the error has been determined for a wide vari- ties on the fatigue resistance [11].
ety of different sample sizes from 3 to 90 spec- In the first place, 15 quasi-static compression
imens. tests were performed at 0.3 MPa/s on the same
The results are presented in graphs that can type of specimens for fatigue tests. The com-
be used for design in cases with a similar order pressive strength fc was 58.9 MPa and the stan-
of magnitude of number of cycles and scatter dard deviation was 3 MPa. For fatigue tests, the
or, in any case, as a reference of the extent of maximum and minimum stress levels were set
the possible error in characterising concrete fa- as 82% and 36% of fc , respectively. Tests were
tigue. With these results, it is possible to define carried out at a load frequency of 10 Hz.
2
José J. Ortega, Gonzalo Ruiz, Rena C. Yu, Nelson Afanador-Garcı́a, Manuel Tarifa, Elisa Poveda and Xiaoxin Zhang
In total, 105 fatigue tests were performed, can be observed, the results cover from an order
with 5 run-out specimens. The run-out limit of 102 to 105 cycles.
was established in 1000000 cycles, around two
or three orders of magnitude higher than the ex- 3 METHOD
pected mean. Those 5 results were discarded for The applied method consists in a sam-
the analysis as they provided incomplete infor- pling technique (‘bootstrap’) where, for differ-
mation. In this manner, the resulting probability ent sample sizes n, values are randomly ex-
distribution is on the safe side. tracted from the set of 100 results in multiple
The probability distribution of the other 100 repetitions. For each sample of values, Eq. 1 is
tests is taken as the real distribution of the ma- fitted and the difference with respect to refer-
terial, taking into account that their number is ence distribution corresponding to the total of
high enough so as not to substantially vary with results is registered. That difference is the error
new tests. Such a distribution is described by made by that particular sample in characterising
means of the probabilistic Weibull model with the fatigue behaviour of the material. The er-
two parameters, defined by Eq. 1. ror is taken as the horizontal distance (in log N )
between both distributions along the cumulative
F (x) = 1 − exp[−(x/λ)κ ] (1) probability axis. The expression that defines the
The variable x is the decimal logarithm of observed error eo in relative terms, at a certain
the number of cycles to failure (log N ), λ is the failure probability p, is:
scale parameter and κ is the shape parameter.
log N (p)sample − log N (p)real
This function is selected due to its suitabil- eo (p) = (2)
log N (p)real
ity to describe phenomena with extreme values
and, in the case of the Weibull distribution with The number of repetitions was determined
two parameters, due to allow the possibility of by a convergence analysis of the difference be-
very low values of N , as observed in real cases. tween the cumulative mean value in log N of
every extracted sample and the mean of the
1.0 original 100 values. The convergence was anal-
Fitted Weibull ysed for a precision degree of 10−10 with sev-
distribution
0.8 Experimental eral sample sizes. The most unfavourable result
results
Failure Probability
3
José J. Ortega, Gonzalo Ruiz, Rena C. Yu, Nelson Afanador-Garcı́a, Manuel Tarifa, Elisa Poveda and Xiaoxin Zhang
1.0 n = 90 n = 40 n=3
Sample size: n 1.0
a) SL = 95 %
0.8 Safety Level: SL n
0.8 3, 5, 7,
9, 11,
Failure Probability
Failure Probability, p
0.6 SL % Error curve 13, 15,
0.6 17, 19,
22, 25,
30, 35,
0.4 p = 0.4 0.4 40, 45,
50, 55,
60, 65,
0.2 0.2
70, 75,
80, 90
ee(n, p, SL)
0.0 0.0
1 2 3 4 5 6 1 10 100
log N Error in log N, ee (%)
n = 90 n = 40 n=3
Figure 2: Error distribution, safety level and expected 1.0
maximum error.
b) SL = 99 %
0.8
Failure Probability, p
4 RESULTS
0.6
In the first place, the scatter of results ob-
tained with samples of different sizes is com- 0.4
pared by two examples (Fig. 3). The overlapped
distributions of all the simulated samples for a 0.2
given number of specimens form a band around
the reference distribution of the total set of val- 0.0
1 10 100
ues. The width of each band shows the observed Error in log N, ee (%)
maximum scatter with respect to it. The dis-
Figure 4: Error curves, ee –p, for the safety levels of a)
tribution obtained by a random sample of that 95% and b) 99%.
number of specimens will be traced within that
band, which can differ more or less from the In the second place, Fig. 4 shows the evolu-
real distribution of the material. tion of the maximum error for each studied sam-
1.0 ple size and for the safety levels of 95% and
a) b) 99%, along the cumulative failure probability.
0.8 n = 7
Failure Probability
4
José J. Ortega, Gonzalo Ruiz, Rena C. Yu, Nelson Afanador-Garcı́a, Manuel Tarifa, Elisa Poveda and Xiaoxin Zhang
sizes that do not reach that error at any point distribution is particularly appropriate for such
of the failure probability range. Likewise, it is a phenomenon. The number of tested speci-
possible to find what curves have an admissible mens is key to the precision of the estimation of
error only at a particular failure probability or the probability distribution followed by the ma-
starting at that point onwards. terial. Due to that high scatter that affects this
Finally, a design curve for fatigue can be de- type of results, the distribution obtained from a
rived with the information of the error curves. random sample of specimens can be very differ-
After performing a given number of tests, the ent from the real distribution of the material if
maximum error is now known for a certain the number of tests is insufficient. This differ-
safety level. However, this error still refers ence is the error made in the estimation of the
to the possibility that the distribution obtained real distribution by that sample.
from the experimental results is to the right of The method developed in this work analyses
the real distribution of the material, which is the maximum committed error depending on
the unsafe side. Therefore, the fatigue design the number of performed tests. After obtaining
curve for that material can be defined by sub- the distribution of the results of 100 tests, which
tracting the corresponding error at each proba- is adopted as the real distribution of the studied
bility p from the experimental distribution. The concrete, a sampling technique has been applied
probability that this resulting curve is to the left, that consists in randomly extracting multiple
or safe side, of the real distribution is equal, by samples from that set of specimens. With each
definition, to the safety level. This design curve sample, the difference in log N with respect to
can be expressed as in Eq. 3. An application ex- the known real distribution is registered. After
ample is shown in Fig. 5. a total of 500000 repetitions, the scatter or error
associated to a determined sample size is found
log N (p)sample
log N (p)design = (3) to follow a normal distribution. From this error
1 + ee (n, p, SL) distribution, it is possible to know the possible
maximum error for a given safety level.
1.0 On the one hand, two examples of scatter
band obtained from different sample sizes are
0.8
shown. For the smaller size, that band is much
Design
Failure Probability
0.6 Real wider than in the other case, that is to say, there
have been random samples that have yielded
0.4 Sample a distribution farther from the real distribution
than when the number of used specimens is
0.2 higher. On the other hand, the error curves of
the 23 studied sample sizes, from 3 to 90 speci-
Tests
0.0 mens, are presented for the safety levels of 95%
1 2 3 4 5 6 and 99%. It is demonstrated how using a re-
log N duced number of specimens the committed er-
Figure 5: Example of fatigue design curve for a safety ror may be very high, which diminishes with
level of 95% with 12 tests. each greater size. These results are useful for
determining the necessary number of tests so
5 CONCLUSIONS that a given admissible error is not surpassed
This work highlights the large scatter of for a certain failure probability and safety level.
loading cycles to failure that can be found in the Furthermore, a design fatigue distribution can
study of concrete fatigue. As a result, fatigue be derived by subtracting from the experimen-
must be characterised by a probabilistic func- tal distribution the possible error corresponding
tion, fitted to experimental results. The Weibull to the required safety level.
5
José J. Ortega, Gonzalo Ruiz, Rena C. Yu, Nelson Afanador-Garcı́a, Manuel Tarifa, Elisa Poveda and Xiaoxin Zhang