Engineering Failure Analysis 18 (2011) 18481857

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Experimental research on the behaviour of high frequency fatigue

in concrete
Youliang Chen a, Jing Ni a,, Ping Zheng a, Rag Azzam b, Youcheng Zhou a, Wei Shao a
a
Department of Civil Engineering, University of Shanghai for Science and Technology, 516 Jungong Road, 200093 Shanghai, PR China
b
Department of Engineering Geology and Hydrogeology, RWTH Aachen University, Lochnerstr. 4-20, Haus A, 52064 Aachen, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Calculating the fatigue strength of concrete under the cyclic load of vehicles when design-
Available online 6 July 2011 ing bridges is an issue which is receiving more and more attention from many engineers
and researchers. Based on this fact, fatigue tests of plain concrete under constant-ampli-
Keywords: tude and stepping-amplitude cyclic loads were conducted. The mechanism which damages
Concrete plain concrete specimens under high frequency fatigue loads was analysed and a non-lin-
High frequency ear accumulative fatigue formula that causes the damage was proposed. A fatigue equation
Cyclic loads
PSN that considers the failure probability p0 was given. The results of this research are a
Fatigue failure
good preparation for further studies into high frequency fatigue tests of concrete cylinders
reinforced with carbon bre.
2011 Elsevier Ltd. All rights reserved.

1. Introduction

Bridges over highways and railways suffer cyclic loads from vehicles which cause the structures to fail, but the failure of
these structures and their components from fatigue loads are far lower than their original strength. Failure from fatigue is
often abrupt and has serious consequences.
Some researchers have already investigated the fatigue performance of concrete. Aas-Jakobsen [1] found there was a good
linear relationship between the logarithm of fatigue life and stress ratio fcmax =fc (fcmax = the maximum value of fatigue com-
pressive stress, fc = the mean value of fatigue compressive stress), and then obtained the Wohler equation in linear form.
Matsushita and Tokumitsu [2] conducted fatigue tests using concrete cylinders. The probability distribution for the fatigue
life of concrete was analysed and the SN (S = stress, N = fatigue life) equation, which includes the lowest fatigue stress level,
was put forward. Hsu [3] introduced a loading speed T into the series of fNR (f = stress level, N = fatigue life, R = correlation
coefcient) equations for fatigue tests on concrete and established the fNRT equation in 3D. He suggested the equations
for high frequency fatigue and low frequency fatigue respectively. Mihashi [4] applied the stochastic process theory when
estimating to the fatigue life of concrete. Liang and Zhou [5] developed a fracture mechanics based fatigue law for asphalt
concrete beams. Based on damage mechanics and tensor theory, Alliche [6] established a three dimensional model for
describing the behaviour of concrete fatigue. Sain and Chandra Kishen [7] concluded that for concrete suffering from fatigue,
a discrete crack may be modelled as an equivalent damage zone where both correspond to the same loss of energy. Hitherto,
many researches have worked on low frequency fatigue in concrete components, whereas research on high frequency fatigue
is relatively recent due to limitations in the test conditions.
In this paper, fatigue tests of plain concrete under constant-amplitude and stepping-amplitude cyclic loads were carried
out. The mechanism that caused the damage was analysed, a non-linear fatigue accumulative damage model was established

Corresponding author.
E-mail addresses: chenyouliang2001@yahoo.com.cn (Y.L. Chen), wendy_1943@163.com (J. Ni).

1350-6307/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2011.06.012
 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857 1849

and a loading frequency correction coefcient that considers how the loading frequency affects the fatigue life was put for-
ward. In addition, probability statistics based on the testing data were analysed and a fatigue equation that considers the
probability of failure was proposed.

2. High frequency fatigue tests of plain concrete

The loading frequency for common fatigue tests is usually below 30 Hz, but can reach 100 Hz for high frequency fatigue
tests. Under high frequency, the time and cost of reaching the fatigue limit and strength of a material are dramatically re-
duced. Here, high frequency fatigue tests with constant amplitude and stepping amplitude were conducted.

2.1. High frequency fatigue tests under constant-amplitude and stepping-amplitude cyclic load

2.1.1. Static loading tests

For these tests, 33 cylindrical C30 concrete specimens, 70 mm in diameter by 100 mm high were prefabricated. They were
manufactured in plastic moulds and were maintained for 28 days at a temperature of 20 3 C before being tested.
From the static loading tests it was found that the ultimate loading capacity of the cylindrical specimens was
Fu = 112.11 kN, so the axial compressive strength is fco = 29.1 MPa. The compressive strength of the cubic samples was
fcu = 36.6 Mpa, and the elastic modulus amounted to Ec = 4.73  104 MPa.

2.1.2. The introduction of a high frequency fatigue testing machine

High frequency fatigue tests were conducted on ZWICK-100HFP5100 test apparatus, manufactured by the ZWICK/ROELL
company, which is shown in Fig. 1. The basic technical parameters of the testing machine include the maximum test load
(100 kN), maximum vibration frequency (150 Hz), maximum static loading capacity (100 kN), and maximum dynamic load-
ing capacity (100 kN 50 kN).

2.1.3. Test schemes

2.1.3.1. Constant-amplitude cyclic loading tests. The test parameters included the biggest stress rmax, the smallest stress rmin,
the mean stress rm, and the stress amplitude ra. The relationship among the parameters is shown in Fig. 2. Table 1 indicates
the loading conditions for the high frequency fatigue tests under constant-amplitude cyclic loads.

Zwick/Roell
620mm
2700mm
1235mm

1040mm
620mm
600mm

530mm 550mm
750mm

(a) (b)
Fig. 1. ZWICK-100HFP5100 test apparatus: (a) ZWICK-100HFP5100; (b) Drawing of ZWICK-100HFP5100 with dimensions.
1850 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857

a

max

m
min

Fig. 2. Schematic diagram for constant-amplitude fatigue test.

Table 1
The loading condition of high frequency fatigue tests.

Group rm/fco ra/fco rmax/fco rmin/fco Number of specimens

Constant-amplitude cyclic loading tests
Group 1 0.45 0.4 0.85 0.05 4
Group 2 0.45 0.35 0.8 0.1 3
Group 3 0.45 0.32 0.77 0.13 3
Group 4 0.45 0.3 0.75 0.15 4
Group 5 0.45 0.28 0.73 0.17 3
Stepping-amplitude cyclic loading tests
Group 1 0.45 0.28 0.73 0.17 4
0.3 0.75 0.15
0.32 0.77 0.13
0.34 0.78 0.11
0.4 0.85 0.05
Group 2 0.45 0.3 0.75 0.15 4
0.35 0.8 0.1

2.1.3.2. Stepping-amplitude cyclic loading tests. Fig. 3 shows the loading form of stepping-amplitude cyclic loading tests and
Table 1 shows the loading conditions.

2.2. Tests results

Constant-amplitude cyclic loading tests were carried out at ve stress levels using 17 specimens while stepping-ampli-
tude cyclic loading tests were divided into two groups using eight specimens. The fatigue life of the specimens is shown in
Table 2. Among the data, two are invalid (the tests were unsuccessful), indicated in bold, three exceeded the limit of fatigue
life of the national norm 2  106, indicated with an underline. The results obtained from the stepping-amplitude cyclic load-
ing test were unsatisfactory and are not listed here.

2.3. High frequency fatigue failure mechanism of the C30 concrete specimens

An analysis of the results of the high frequency fatigue tests showed there are two main modes of fatigue failure:

(1) Vertical failure. During loading it was noted that the loading plates of the fatigue testing machine were the same length
as the end planes of the concrete specimen. Horizontal fatigue tensile stress will cause vertical micro-cracks to begin
and then develop into vertical macro-cracks which will cause the specimen to rupture. The vertical failure mode is
shown in Fig. 4a.
(2) Conical failure. Cracks in the concrete specimens did not develop along the loading direction during loading. The nal
rupture macro-cracks look like cones. The conical failure mode is shown in Fig. 4b.

Fig. 3. Schematic diagram for stepping-amplitude fatigue test.

 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857 1851

Table 2
Fatigue life results of high frequency fatigue tests with constant amplitude.

Specimen rmax/fco rmin/fco Fatigue life Mean fatigue life

Group 1
PCF15 0.85 0.059 10,331 26,710
PCF16 944
PCF19 14,708
PCF24 55,090
Group 2
PCF13 0.8 0.125 61,063 59,518
PCF17 41,773
PCF20 75,718
Group 3
PCF5 0.77 0.169 244,165 248,550
PCF21 252,934
PCF22 3870
Group 4
PCF6 0.75 0.200 804,676 15,15,514
PCF12 21,00,011
PCF18 21,00,000
PCF23 10,57,368
Group 5
PCF7 0.73 0.233 21,00,003 14,05,245
PCF8 940,873
PCF9 11,74,859

Fig. 4. Fatigue failure mode under constant-amplitude cyclic loads: (a) Vertical failure; (b) Conical failure.

The results of fatigue testing showed that vertical failure was the main failure mode. With the stepping-amplitude cyclic
loading tests, because vibration occurred over a long period of time, surface cracks and internal cracks were fully developed
and some specimens were crushed. Fig. 5 shows how specimen PCF21 failed in the constant-amplitude cyclic loading test.

3. The non-linear accumulative high frequency fatigue damage formula of C30 plain concrete

Based on the results from axial fatigue testing of high strength concrete, Wu et al. [8] modied the linear accumulative
fatigue damage formula by considering the discreteness of the fatigue tests and the order of loading. He put forward the fol-
lowing formula for computing the accumulated fatigue damage and predicting the residual fatigue life:
1852 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857

Fig. 4 (continued)

Fig. 5. Crack development of specimen PCF21.

X ni
D c1 c2  1 1
i
Nfi

where c1 and c2 are the correction coefcients that consider the discreteness of fatigue tests and the order of loading respec-
tively. N is the fatigue life and ni is the cyclic loading number, both of which correspond to the stress level i.
Cao [9] replaced c1c2 with the function si(ni), and gave the accumulative damage formula at the stress level i after ni cycles
of loading:
ni
Dt si ni 2
Nfi
Fig. 6 shows si(ni)  ni curves for concrete specimens in the constant-amplitude cyclic loading tests at ve stress levels.
The following equations were obtained through curve tting:

Smax 0:73 : sn 2:1154n0:988 ; R 0:9997;

 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857 1853

Fig. 6. sn Curves under constant-amplitude cyclic loads at ve stress levels.

Smax 0:75 : sn 1:1503n1:032 ; R 0:9971;

Smax 0:77 : sn 2:8495n1:101 ; R 0:9915;

Smax 0:80 : sn 6:2902n0:988 ; R 0:9985;

Smax 0:85 : sn 1:5427n1:04 ; R 0:9994;

1854 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857

In the equations listed above, Smax is the maximum stress level and R is the correlation coefcient.
Five of these equations could be unied as:

si n ai nbi 3
where ai and bi are material properties corresponding to the stress level i, which can be obtained by constant-amplitude cyc-
lic loading tests.
Hence, the following accumulative damage formula is suggested.
X ni X b ni
D si ni ai ni i  4
i
Nfi i
Nfi

Eq. (4) is not only suitable for computing the fatigue damage under constant-amplitude cyclic loads, but for analysing the
development of fatigue damage under stepping-amplitude cyclic loads.

4. Basquin functions

Fig. 7 shows lg Slg Nf curves obtained from normal low frequency fatigue tests and high frequency fatigue tests with con-
stant-amplitude cyclic loads respectively. The data from low frequency fatigue tests were obtained from the Ref. [9] where
the compressive strength of concrete was 20.47 MPa.
When compared with data from the low frequency fatigue tests, the fatigue life in high frequency fatigue tests is longer at
the same stress level, which reects how the loading frequency affects the fatigue performance of concrete.
In 1910, Basquin established the famous Basquin equation for constant-amplitude cyclic loading tests which shows the
relationship between stress amplitude and fatigue life:

ra r0f 2Nf b 5

where ra is the stress amplitude, Nf is the fatigue life corresponding to the stress amplitude ra, b is Basquin index, and r0f is
the coefcient of fatigue strength. The Basquin equation is also suitable for describing a raNf relationship in high frequency
fatigue tests.
By making a logarithmic transformation in both sides of the Eq. (5):
lg ra lg r0f b lg 2Nf 6

The Basquin equations for the curves in Fig. 7 are:

High frequency : ra 0:774587  2Nf 0:06597 7

Low frequency : ra 0:426177  2Nf 0:03455 8

The fatigue limit ra corresponding to a different cyclic loading number can be calculated from Eqs. (7) and (8).

5. Corrected coefcients for high frequency fatigue tests

According to the Chinese national norm, N = 2  106 is used as the limit for the low frequency fatigue life of concrete
when designing a bridge over a highway.

Fig. 7. lg S lg N f curves under high frequency and low frequency fatigue loads.
 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857 1855

From a comparison between the results of low frequency and high frequency fatigue tests in Fig. 7, it can be seen that high
frequency can improve the fatigue characteristics of concrete. When the results of high frequency fatigue test are applied to
solve the practical low frequency fatigue problems, the effect of high loading frequency should be considered.

5.1. The denition of corrected coefcient

The high frequency corrected coefcient is dened as the ratio of fatigue limit raL under low frequency to the fatigue limit
raH under high frequency:
raL
/ 9
raH
From Eqs. (5) and (9), the following equation is obtained:

/ /0 2Nf Db 10

where

r0fL
/0
r0fH

Db b L  b H

u0 is the ratio of the strength coefcient of a low frequency fatigue test to a high frequency fatigue test, while Db is the dif-
ference between the strength index of a low frequency fatigue test and a high frequency fatigue test.

5.2. The correction of high frequency fatigue test results

If the corrected high frequency coefcient u is known, then the fatigue characteristics of concrete under low frequency
loading could be obtained from the results of a high frequency fatigue test through the following equation:

raL /r0fH 2Nf bH 11

Table 3 shows the corrected high frequency coefcient u corresponding to different fatigue life from 1  10 to 2  106, 4

which is obtained from high and low frequency fatigue tests.

Table 3
High frequency and low frequency fatigue limits along with high frequency correction coefcient corresponding to various fatigue cycles.

Cycle High frequency fatigue limit Low frequency fatigue limit High frequency correction coefcient, /
4
1  10 0.403 0.303 0.868
1  105 0.346 0.280 0.887
1  106 0.297 0.258 0.807
2  106 0.284 0.252 0.751

Fig. 8. lg ra lg 2N f curves obtained from modied curves of high frequency fatigue tests and low frequency fatigue tests.
1856 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857

Fig. 8 shows two lg ra lg 2N f curves obtained from modied results of a high frequency fatigue test and the results of a
low frequency fatigue test respectively. Two curves are congruent so it is feasible to calculate the results of concrete fatigue
under a low frequency condition rather than from high frequency fatigue tests.

6. Fatigue curves and fatigue equation

6.1. Fatigue curves

S lg N data points from the high frequency fatigue tests and the corresponding tting curves are shown in Fig. 9. S is the
stress level and N is the corresponding fatigue life.

6.2. PSN fatigue equation considering failure probability

1
 g
For the given failure probability p0 , the equivalent fatigue life N  j ln1  p0 jb can be calculated. The results are listed in
Table 4. Table 5 shows the regression coefcients c, d and the correlated coefcient r that corresponds to various fatigue fail-
ure probability p0 , which could be obtained from a linear regression of the data in Table 4 using lg S lg d  c1  R2 lg N.
p0 could be determined from the required reliable possibility, and then c and d could be received from Table 5. The PSN
equation can be obtained by substituting c and d into lg S lg d  c1  R2 lg N.

Fig. 9. S lg N data points from the high frequency fatigue tests and the corresponding tting curves.

Table 4
Equivalent fatigue life corresponding to failure probability p0 .

Failure probability, p0 Stress level, S

0.85 0.80 0.77 0.75 0.73
0.05 348 18,562 161,234 179,158 152,066
0.1 1148 24,481 164,712 273,516 224,914
0.2 3986 32,669 168,418 425,147 338,242
0.3 8677 39,124 170,776 560,093 436,497
0.4 15,744 44,918 172,605 691,754 530,647
0.5 26,119 50,511 174,174 827,678 626,441

Table 5
Regression coefcients c and d along with correlation coefcient r, corresponding to various fatigue failure probability p0 .

Failure probability, p0 Regression coefcient, d Regression coefcient, c Correlation coefcient, r

0.05 0.9645 0.0212 0.9331
0.1 1.0158 0.0252 0.9557
0.2 1.0882 0.0304 0.9734
0.3 1.1425 0.0339 0.9749
0.4 1.1849 0.0364 0.9656
0.5 1.2162 0.0380 0.9467
 Y. Chen et al. / Engineering Failure Analysis 18 (2011) 18481857 1857

The fatigue equation with a failure probability of 5% is:

lg S 0:01568  0:02121  R2 lg N 12
The fatigue equation with a failure probability of 50% is:

lg S 0:08502  0:0381  R2 lg N 13

7. Conclusions

Through the experimental research work and theoretical analysis, the following conclusions can be drawn:

(1) The two main failure modes of plain concrete in high frequency fatigue tests are vertical failure and conical failure.
(2) The results of low frequency fatigue tests can be calculated from the results of high frequency fatigue tests using the
corrected high frequency coefcient u.
(3) It is more reasonable to establish a PSN fatigue equation that corresponds to a certain failure probability.
(4) Research work contributes to a discussion of the characteristics of a bre reinforced concrete column in high fre-
quency fatigue tests.

Acknowledgments

The nancial support from National Natural Science Foundation of China (10872133), the Pujiang Talent Program
(09PJ1407700) and Innovation Program of Shanghai Municipal Education Commission (08YZ91) for this study is gratefully
acknowledged.

References

[1] Aas-Jakobsen K. Fatigue of concrete beams and columns. NTH Inst Betonkonstruksjoner Bull 1970;70(1):148.
[2] Matsushita H, Tokumitsu Y. A study on compressive fatigue strength of concrete considered survival probability. Proc JSCE 1972;198:12738.
[3] Hsu TCT. Fatigue of plain concrete. ACI Mater J 1981;78(8):292305.
[4] Mihashi H. Stochastic approach to study fatigue of concrete. Eng Fract Mech 1987;28(56):78593.
[5] Liang YR, Zhou J. Prediction of fatigue life of asphalt concrete beams. Int J Fatigue 1997;19(2):11724.
[6] Alliche A. Damage model for fatigue loading of concrete. Int J Fatigue 2004;26(9):91521.
[7] Sain T, Chandra Kishen JM. Energy-based equivalence between damage and fracture in concrete under fatigue. Eng Fract Mech 2007;74(15):232033.
[8] Wu PG, Zhao GY, Bai ML. Fatigue behavior of high strength concrete under compressive cyclic loading. China Civ Eng J 1994;27(3):3340 [in Chinese].
[9] Cao W. Experimental and theoretical research on fatigue properties of plain concrete under triaxial cyclic loading with constant lateral pressure. In:
Doctoral dissertation for Dalian University of Technology, Dalian, China; 2004 [in Chinese].