You are on page 1of 11

Articles

Ludger Lohaus DOI: 10.1002/suco.201100054


Nadja Oneschkow*
Maik Wefer

Design model for the fatigue behaviour


of normal-strength, high-strength and
ultra-high-strength concrete
Fatigue design according to CEB-FIP Model Code 90 [1] is limited purpose of comparison. Based on these test results, new
to concrete grades up to C80. In addition, the design rules include Woehler curves and a new approach to the fatigue design
a strength-dependent reduction in the fatigue reference strength, of normal-strength, high-strength and ultra-high-strength
which leads to uneconomical design of high-strength concrete. concrete under uniaxial compressive fatigue loading was
Considering comprehensive knowledge now available concern- developed in [2], and subsequently developed further [3].
ing the fatigue behaviour of normal-strength and high-strength Therefore, the fatigue design values of CEB-FIP Model
concretes, the amount of this reduction can no longer be justi- Code 90 [1] were modified in such a way that they apply to
fied. A new design model for compressive fatigue loading and its normal-strength, high-strength and, additionally, ultra-
derivation is presented in this article [2, 3]. A comparison be- high-strength concrete. The design model and its capabili-
tween the new design model and the current standard ones [1] ties are presented in this paper. A comparison between the
reveals that the new design model ensures safe and economical
new design model and the design Model according to
design of normal-strength, high-strength and ultra-high-strength
CEB-FIP Model Code 90 [1] shows that the new design
concrete. This new design model is included in the new fib Model
model ensures safe and economical design of concrete un-
Code 2010 [4].
der fatigue loading. The design model presented is includ-
Keywords: fatigue, design model, high-performance concrete, ed in the new fib Model Code 2010 [4].
Model Code 2010
2 Experimental investigations
1 Introduction 2.1 Test specimens

The design of structures with regard to the material prop- The following results were obtained based on experimen-
erties of high-strength and ultra-high-strength concretes tal investigations on two ultra-high-strength concrete mix-
(UHPC) usually leads to lightweight and slender structural tures, which were the standard mixtures in priority
elements that differ significantly from the traditional bulky programme 1182 Sustainable Building with Ultra-High-
structures. Concurrently, those structures with a reduced Performance Concrete funded by the German Research
dead weight are more susceptible to variable loads, e.g. Foundation (DFG).
non-static loading. As a consequence, the influence of The ultra-high-strength fine-grained concrete (M2Q),
cyclic stresses increases and fatigue becomes relevant for with a maximum grain size of 0.5 mm, and the coarse-
such concrete structures. The Woehler curves of CEB-FIP grained concrete (B4Q), with a maximum grain size
Model Code 90 [1] are based on research conducted by of 8.0 mm, had 28-day compressive strengths of
Petkovic [5, 6]. However, types of concrete with compres- fc,cube,100 = 160MPa and fc,cube,100 = 180MPa respectively,
sive strengths > 100MPa were not considered in that re- both after storage underwater [7]. Both mixtures con-
search work, reflecting the fact that there was no reliable tained 2.5 % by vol. of high-strength smooth steel fibres
experience with such types of concrete in those days. Fur- with a length of 9.0 mm and an l/d ratio of 60. The exper-
thermore, that research was conducted based on a limited imental investigations were carried out using cylindrical
number of samples [5, 6]. test specimens with dimensions d x h = 60 180mm. The
Hence, one focus of the research work presented in formwork was removed after 48 hours and the specimens
this paper is an ultra-high-strength concrete (UHPC) with were then heat-treated at 120 C for 2 days. Following the
an average compressive strength of 180MPa [2]. The num- heat treatment, the cylinders were stored in standardized
ber of samples used in these fatigue tests was much larger conditions (20 C, 65 %RH) until testing. Before testing,
than Petkovics. Additional investigations were conducted the loaded surfaces of the specimens were ground parallel
on a normal-strength and a high-strength concrete for the and then polished.

* Corresponding author: n.oneschkow@baustoff.uni-hannover.de 2.2 Testing machines


Submitted for review: 18 November 2011
Revised: 22 March 2012 Most of the fatigue tests, with expected numbers of cycles
Accepted for publication: 29 April 2012 to failure of up to N = 2.0106, were conducted using a

182 2012 Ernst & Sohn Verlag fr Architektur und technische Wissenschaften GmbH & Co. KG, Berlin Structural Concrete 13 (2012), No. 3
L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

servo-hydraulic universal testing machine with a test fre- log N = 24.9 Sc,max + 25.4; B = 0.78
quency of fP = 10Hz. A 1MN actuator was used for these for Sc,min = 0.40 (3)
tests. In cases where the expected numbers of cycles to
failure were higher than N = 2.0106, the tests were car- In accordance with [10], an investigation was carried out
ried out using a resonance testing machine with a test fre- to discover whether the variances in the number of cycles
quency of approx. fP 60Hz. Both testing machines were to failure in the fatigue tests can be traced back to the vari-
used for expected numbers of cycles to failure of between ances in static compressive strength. Therefore, a normal
N = 5.0105 and N = 2.0106. The test results show that distribution was used to describe the static compressive
compared with the servo-hydraulic testing machine, the strength and the logarithmic numbers of cycles to failure.
characteristics of the resonance testing machine led to The tolerance range, which results from the variances in
lower numbers of cycles to failure [2]. Thus, the considera- the static compressive strength, can be determined for the
tion of these test results is on the safe side. 95 % level of confidence. The tolerance range is exemplar-
ily calculated for a maximum compressive stress level of
2.3 Fatigue tests and evaluation of test results Sc,max = 0.70 according to [11] as follows [2]:

Sufficient data are required for determining reliable re- (k T s)


TstU = Sc,max 0.61 < Sc,max =
gression lines in the finite life fatigue strength range. Sc,max
Therefore, single-level Woehler fatigue tests were carried (k T s)
out on the two types of ultra-high-strength concrete with = 0.70 < 0.79 Sc,max + = TstO (4)
Sc,max
constant minimum compressive stress levels of
Sc,min = 0.05, Sc,min = 0.20 and Sc,min = 0.40. The mean val- where
ues of static compressive strengths fcm,i were determined U; T O
Tst lower and upper bounds of tolerance range,
st
using at least three specimens from the same batch and which results from variances in the static com-
with the same geometry. These mean values were used as pressive strength
reference strengths for calculating Sc,min and Sc,max. The
x static compressive strength
st
regression lines shown in Fig. 1 were established based on kT value for determining the tolerance range
88 tests with a minimum compressive stress level of
Sc,min = 0.05, 21 tests with Sc,min = 0.20 and 12 tests with As an example, Fig. 2 compares the tolerance range (Tst U;

Sc,min = 0.40. O
Tst ) resulting from the variances in static compressive
No significant differences in the regression lines strength (x ) with the tolerance range resulting from the
st
were found in the test results for UHPCs with different variances in number of cycles to failure for Sc,max = 0.70
grain compositions. This confirms the findings in [8] and and Sc,min = 0.05.
[9] that different grain compositions have no significant ef- It is obvious that the range of variation (Tst U

fect on the fatigue behaviour. Thus, the two UHPC mix-


xst Tst ) resulting from the variances in the static com-
O

tures were not considered separately. Taking into account pressive strength, and thus included in the fatigue tests,
the assumptions mentioned above, the regression analysis completely covers the tolerance range of the regression
based on single values of the logarithmic number of cycles lines. Accordingly, for ultra-high-strength concrete under
to failure results in the following regression lines for ultra- fatigue loading, similarly to normal and lightweight con-
high-strength concrete (see also Fig. 1): crete [10], no additional variances could be observed due
to fatigue loading.
log N = 12.4 Sc,max + 14.1; B = 0,87 The influence of the minimum stress level Sc,min
for Sc,min = 0.05 (1) must be taken into account for a complete description of

log N = 18.7 Sc,max + 19.2; B = 0.96 1.0


for Sc,min = 0.20 (2)
0.9
Sc,max = c,max / fcm,i [-]

1.0
T Ost = xst + kT * s
0.8
Sc,max = c,max / fcm,i [-]

0.9
x st
0.7
0.8 T Ust = xst - kT * s
0.6
0.7
single value log N
0.5
regression line
0.6 Sc,min = 0.05
Sc,min = 0.20 range of tolerance Sc,min = 0.05
Sc,min = 0.40 0.4
0.5 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 number of cycles to failure log N [-]
number of cycles to failure log N [-]
Fig. 2. Comparison between range of variation in static strength and toler-
Fig. 1. Experimental test results and regression lines ance range for number of cycles to failure

Structural Concrete 13 (2012), No. 3 183


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

1.0

0.8
Sc,max = c,max / fcm,i [-]

0.6

Sc,max
0.4

0.2 boundary line for UHPC for N = 2.0 x 106


boundary line for N = 2.0 x 106 from Petkovic
boundary line for minimum stress level
0.0
0.0 0.2 0.4 0.6 0.8 1.0 Sc,min [-]
log N [-]
Sc,min = c,min / fcm,i [-]

Fig. 3. Comparison between fatigue strengths of our investigations and of [5] Fig. 4. Regression lines shown in a 3D Model

the fatigue behaviour because an increasing minimum knowledge gained from the literature, the derivation takes
stress level Sc,min in combination with a constant maxi- the following requirements into consideration [2]:
mum stress level Sc,max leads to higher numbers of cycles Intersection with the ordinate at Sc,max = 1.0 [5, 6]
to failure. The results of the regression analysis are com- Linear approximation, at least for the range up to log
pared with the investigations by Petkovic [5] in a Good- N = 7 [12]
man diagram for a number of cycles to failure of Deviation from test results that is on the safe side
N = 2.0106 (see Fig. 3). The investigations of Petkovic [5] Connection to high numbers of load cycles (N > 107)
are the basis for the S-N relations in CEB-FIP Model Code and thus continuous description for all maximum stress
90 [1]. levels Sc,max
For the UHPC tested, it is obvious that higher An asymptotic approach to the respective minimum
permissible amplitudes were reached than those accord- stress levels for high numbers of load cycles [1]
ing to [5] for all minimum compressive stress levels
Sc,min,i. This means that the fatigue strength of the UHPC 3.2 Mathematical description of failure surface
specimens investigated is underestimated by the S-N rela-
tions according to CEB-FIP Model Code 90 [1]. According to the requirements described above, it is possi-
In addition to the test series described above, it was ble to define the parameterized failure surface for uniaxial
also necessary to analyse different influencing factors rele- compressive fatigue. A three-dimensional system of coor-
vant to the fatigue behaviour of UHPCs [2]. These further dinates with axes Sc,max, Sc,min and logN, according to [2],
investigations mainly focused on the influence of steel fi- is therefore considered (Fig. 5). This system requires the
bres and heat treatment. The results of these investiga- four fatigue strength values in the plane log N2 = const.
tions are included in Fig. 10, but are not incorporated in
detail here. P1,fat = (log N2, Sc,max(log N2, Sc,min,1), Sc,min,1),

3 Development of a material model for uniaxial P2,fat = (log N2, Sc,max(log N2, Sc,min,2), Sc,min,2),
fatigue loading
3.1 Requirements for the mathematical description P3,fat = (log N2, Sc,max(log N2, Sc,min,3), Sc,min,3) and

The material model describing the behaviour of concrete P4,fat = (log N2, Sc,max(log N2, Sc,min,4), Sc,min,4)
under uniaxial fatigue loading has been developed based
on the experimental investigations, taking the minimum to be determined.
compressive stress level into consideration. The material The points P1,fat, P2,fat, P3,fat and P4,fat represent four
model is determined by combining the Woehler and fatigue strengths determined for a previously defined num-
Goodman diagrams (Fig. 4). The main objectives for de- ber of cycles to failure log N2 based on experimental in-
veloping this model were a complete description of the fa- vestigations or by extrapolating the regression lines. Each
tigue behaviour and thus the definition of a failure surface fatigue strength is specified by the maximum stress level
for uniaxial compressive fatigue loading. Sc,max,i, the minimum stress level Sc,min,i and the number of
Based on the comparative research conducted on nor- cycles to failure log N2. For high log N2 and minimum
mal-strength and ultra-high-strength concrete, and based on stress levels in the range of the sustained static loading

184 Structural Concrete 13 (2012), No. 3


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

Sc,max,i Equation 7
1.0 Pi,0

P4,fat Equation 15
Sc,min,i = const.

Pi,fat
P3,fat
Sc,max [-]

P2,fat
0.0
log N2 log N
P1,fat
Fig. 6. Woehler curve shown in the plane Sc,min,i = const.

log N2 = const.
Pi,fat = (N2, Sc,max(log N2), Sc,min)
Sc,min
log N [-]

as the first parameter and takes the gradient at this point


as the second parameter (Fig. 6). The approach for this ex-
Fig. 5. Input parameters required for the material model (3D view)
ponential function in the logarithmic space is

Sc,max Sc,min = a N = a e ln x = a 10 log(x) = a 10 x


(

= f log x = f(x) ) (8)
strength, point P4,fat is equal to the sustained static loading
strength of the concrete investigated (Sc,max = Sc,min). where a, und N > 0.
Firstly, the plane Sc,min,i = const. is considered (see The following requirements have to be met when de-
Fig. 5). This is correlated with a Woehler curve for an ar- termining parameters a and mentioned above:
bitrary minimum compressive stress level Sc,min,i. In this
plane, the points ) = f(N ) = a 10 x 2
f(x = Sc,max(log N2 ) Sc,min (9)
2 2

Pi,0 = (0.0; 1.0) and f(x 2 ) = a 10 x 2 ln(10) = m (10)

Pi,fat = (log N2, Sc,max(log N2)) where x 2 = log N2 and m is taken from Eq. (6). Divid-
ing Eq. (10) by Eq. (9), parameter is obtained as
are already known. Pi,0 is the starting point of the Woehler follows:
curve at Sc,max,i = 1.0 [5, 6] and Pi,fat represents the fatigue
strength at logN = logN2. A function Sc,max (log N) with a 10 x 2 ln(10) m
= ln(10) = (11)
logN 0 is considered in this plane. A linear equation is a 10 x 2 S c,max(log N2 ) Sc,min
formulated for 0 log N log N2 which passes through
points Pi,0 and Pi,fat, see Fig. 6. Substituting both points in- m
= (12)
to the linear equation presented by ln(10) Sc,max(log N2 ) Sc,min

y = Sc,max (log N) = m x + b = m log N + b (5) Substituting in Eq. (9), the required parameter a is
calculated as:
means it is possible to calculate the variables

mlog N2

Sc,max(log N2 ) 1 Sc,max(log N2 ) 1 ln(10)Sc,max(log N2 )Sc,min
m= = (6) a = Sc,max(log N2 ) Sc,min 10 (13)
log N2 0 log N2

and b = 1 and thus the straight line Inserting parameter and a into Eq. (8) then results
in:
S (log N2 ) 1
Sc,max(log N) = c,max log N + 1 (7)
log N2 Sc,max(log N) = Sc,max(log N2 ) Sc,min
m (log N log N2 )
This equation clearly defines the Woehler curve in the ln(10) Sc,max(log N2 ) Sc,min
range up to logN = logN2 for all minimum compressive 10 + Sc,min (14)
stress levels Sc,min,i. An exponential function is chosen for
numbers of cycles to failure logN > logN2 which asymp- This function is a monotonic decreasing exponential func-
totically approaches the minimum compressive stress lev- tion. It continues the line expressed by Eq. (7) for the area
el value of the respective regression line. In the non-loga- 0 logN logN2 at the point where logN = logN2 for
rithmic space, this function intersects the point logN > logN2.

Structural Concrete 13 (2012), No. 3 185


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

Solving the equation for logN leads to: 1


b=
det
log N = log N2 + x x x y y y
23 34 42 2 3 4
Sc,max Sc,min + y1 x 21x 31x 23y 2y 3 + x 31x 41x 34 y 3y 4 + x 41x 21x 42y 4 y 2
ln(10) Sc,max(log N2 ) Sc,min log
Sc,max(log N2 ) Sc,min (18)
Sc,max(log N2 ) 1 1
c=
log N2 det
(15) (( ) ( ) ( )
x 23 x12 y 3c(21)(41) + x 24 x12 y 4c(31)(21) + x 22 x12 y 2c(41)(31) )
(19)
Eq. (15) defines the Woehler curves for all maximum com-
pressive stress levels 0< Sc,max 1 and for each minimum
compressive stress level Sc,min,i (see Fig. 6).
d =
1
(
x y c +x y c +x y c
det 21 2 (41)(31) 31 3 (21)(41) 41 4 (31)(21) ) (20)
Now consider the plane log N = log N2 = const.
(Fig. 7). This plane represents the modified Goodman dia- where
gram for the number of cycles logN = logN2, which de- det = x 2x 3x 32x 41y 2y 3 + x 3x 4 x 43x 21y 3y 4 + x 4 x 2x 24 x 31y 4 y 2
notes the change from the linear to the non-linear part of + x1x 2x 21x 43y1y 2 + x1x 3x 31x 24 y1y 3 + x1x 4 x 41x 32y1y 4
the Woehler curve. The discontinuous rational function (21)
a + b Sc,min where xij = xi xj, Yij = yi yj and c(i1)(j1) = xi1yj1 xj1yi1.
Sc,max = (16)
1 + c Sc,min + d Sc,min
2 (22)

was chosen to approximate the progression of the Good- Based on these considerations and the combination of the
man diagram. This function includes the following four fa- approximations described, two partial planes that depend
tigue strengths: on log N, Sc,max and Sc,min are generated in the three-di-
mensional system of coordinates (Fig. 5).
P1,fat = (Sc,min,1, Sc,max (Sc,min,1)) = (x1, y1), The partial plane in the subspace logN logN2 is:

P2,fat = (Sc,min,2, Sc,max (Sc,min,2)) = (x2, y2), S (log N2 ) 1


Sc,max(log N) = c,max log N + 1
log N2
P3,fat = (Sc,min,3, Sc,max (Sc,min,3)) = (x3, y3) and
a + b Sc,min
1
P4,fat = (Sc,min,4, Sc,max (Sc,min,4)) = (x4, y4). 1 + c Sc,min + d Sc,min
2
= log N + 1
(23)
log N2
It should be noticed here that 0< x1 < x2 < x3 < x4. The

four parameters a, b, c and d required are determined
depending on the points P1,fat, P2,fat, P3,fat, and P4,fat [2]:
Solving for log N:
1
a=
det
y x x x x y y + x x x x y y + x x x x y y log N =
(S c,max )
1 log N2
1 2 31 41 43 3 4 3 41 21 24 4 2 4 21 31 32 2 3 a + b Sc,min
x1x 23x 34 x 42y 2y 3y 4 1 (24)
1 + c Sc,min + d Sc,min
2
(17)

S c,max,i In the subspace log N > log N2, the partial area is ex-
pressed by:

Sc,max(log N) = Sc,min + Sc,max(log N2 ) Sc,min


1.0
P 4,fat
P 3,fat m(log N log N2 )
0.8 ln(10)Sc,max(log N2 )Sc,min
P 2,fat 10
a + b Sc,min
0.6 = Sc,min + S
c,min
1 + c Sc,min + d Sc,min
2

P 1,fat m(log N log N2 )
0.4 a + bSc,min
ln(10) Sc,min
1+ cS + d S 2

10 c,min c,min
(25)
0.2
where
a + b Sc,min
S c,min,i 1
Sc,max(log N2 ) 1 1 + c Sc,min + d Sc,min
0.0 0.2 0.4 0.6 0.8 1.0 2
m= = (26)
Fig. 7. Fatigue strengths Pi,fat shown in the plane logN = logN2 = const. log N2 log N2

186 Structural Concrete 13 (2012), No. 3


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

Using a, b, c and d according to Eqs. (17) to (20) and solv-


ing for logN results in the following equation:
a + b Sc,min
log N = log N2 + ln(10) Sc,min
1 + c Sc,min + d Sc,min
2 P 4,fat



Sc,max Sc,min

S c,max [-]
log P 3,fat

a + b Sc,min
Sc,min P 2,fat
1 + c Sc,min + d Sc,min
2
log N2 P 1,fat
(27)
a + b Sc,min
1
1 + c Sc,min + d Sc,min
2

Applying Eqs. (24) and (27), the failure surface for uniaxi- log N2 = const. S c,min [-]
log N [-]
al compressive fatigue loading can be unequivocally gen-
erated for arbitrary maximum and minimum stress levels
using four certain fatigue strengths (P1,fat, P2,fat, P3,fat and
Fig. 8. Failure surface for uniaxial fatigue loading
P4,fat). These fatigue strengths have to be determined ac-
cording to experimental test results for the same ultimate
number of cycles to failure. It is assumed that the sustained static loading strength can
be set to Sc,max = Sc,min = 0.85, referring to the characteris-
3.3 Determining the model parameters tic compressive strength [2]. Therefore, the fatigue
strength P4,fat is
According to the above considerations and taking into ac-
count the investigations conducted on ultra-high-strength P4,fat: Sc,max,4 (Sc,min,4 = 0.85, log N = 8) 0.85
concrete, the following input quantities are established for
the material model. The starting point for the regression The discontinuous rational function according to Eq. (16)
lines for all minimum compressive stress levels is located is used to approximate the progression of the Goodman
at Sc,max = 1.0 [6]. Investigations by Klausen [12] show that diagram. The parameters a, b, c and d are calculated as
a linear correlation between the number of cycles to fail- a = 0.45, b = 1.8, c = 1.8 and d = 0.3 by inserting the val-
ure and the maximum compressive stress level can be as- ues for the fatigue strengths P1,fat, P2,fat, P3,fat and P4,fat in
sumed up to a number of cycles to failure of N = 108. Ac- Eqs. (17) to (22). Substituting the parameters a, b, c and d
cording to this, a linear extrapolation of the regression in Eqs. (24) and (27) clearly defines the fatigue failure sur-
lines detected experimentally was carried out up to face for uniaxial fatigue loading for arbitrary maximum
logN = 8. A kink in the Woehler curve, as presented in [5], and minimum compressive stress levels (Fig. 8).
cannot be deduced from the investigations conducted.
Thus, the material model provides a description of the ma- 3.4 Verification of the material model developed
terial behaviour that is on the safe side. The lines for
0 logN logN2 = 8 result as follows: Experimental test results for different concrete types un-
der compressive fatigue loading are shown in Fig. 9, to-
log N = 15.87 (1.0 Sc,max) for Sc,min = 0.05 (28) gether with the Woehler curve according to CEB-FIP Mod-
el Code 90 [1] and the new Woehler curve. Therefore, a
log N = 19.99 (1.0 Sc,max) for Sc,min = 0.20 (29) total of 272 experimental tests were analysed, presenting
the numbers of cycles to failure as mean values. The max-
log N = 26.42 (1.0 Sc,max) for Sc,min = 0.40 (30) imum and minimum compressive stress levels were calcu-
lated using the mean values of the static short-term
The transition point between linear and non-linear parts strength as the reference strength. Each mean value was
of the Woehler curve is set to be logN = logN = 8. Further previously determined using at least three specimens. Nor-
input quantities are the maximum compressive stress level mal-strength, high-strength and ultra-high-strength con-
Sc,max,i (log N = 8) and the minimum compressive stress cretes with and without fibres, as well as with and without
level Sc,min,i (logN = 8) of the fatigue strengths P1,fat, P2,fat, heat treatment, were investigated using different test fre-
P3,fat and P4,fat. For logN2 = 8, they are calculated using quencies. The minimum compressive stress level was
the extrapolated regression lines considering the respec- Sc,min = 0.05 in each test.
tive minimum stress levels Sc,min,i = 0.05, 0.20,0.40: It can be seen that the numbers of cycles to failure
are quite similar for all types of concrete investigated. Fur-
P1,fat: Sc,max,1 (Sc,min,1 = 0.05, log N = 8) 0.50 thermore, the numbers of cycles to failure for all concrete
types are higher than those according to [1]. It is impor-
P2,fat: Sc,max,2 (Sc,min,2 = 0.20, log N = 8) 0.60 tant to mention that the Woehler curve [1] is rather con-
servative. The same conclusion can be drawn regarding
P3,fat: Sc,max,3 (Sc,min,3 = 0.40, log N = 8) 0.70 concretes C80, C70 without fibres and C90 with fibres, al-

Structural Concrete 13 (2012), No. 3 187


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

1.0 1.0
ntests = 272 Run out
0.9
0.9 Run out
maximum stress level S c,max [-]

maximum stress level S c,max [-]


0.8
0.8 0.7

0.7 0.6
0.5
0.6
0.4
0.5 0.3
0.2
0.4
S c,min = 0.05 0.1
0.3 S c,min = 0.05
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0
number of cycles to failure log N [-]
number of cycles to failure log N [-]
Woehler curve Model Code 90 [1] Woehler curve Model Code 90 [1]
New Woehler curve Hannover New Woehler curve Hannover for NC, HP C, UHP C
mean values C20 without fibres, 10 Hz [2] (with and without fibres)
mean value C70 without fibres, 0.5 Hz Approach for NC of Hsu (without fibres) [17]
mean values C80 without fibres, 10 Hz [13] Regression for UHP C of University Aalborg (without fibres) [15]
mean values C80 without fibres, 5 Hz [13] Regression for UHP C of University Kassel (with fibres) [16]
mean values C80 without fibres, 1 Hz [13] Regression for NC from Klausen (without fibres) [12]
mean values C80 without fibres, 0.1 Hz [13]
mean value C90 with fibres, 0.5 Hz Fig. 10. Comparison between the new Woehler curves and data taken from
mean values C150 without fibres, 10 Hz [14] the literature
mean values C200 with fibres, heat-treated, 10 Hz [2]
mean values C200 without fibres, heat-treated, 10 Hz [2] normal-strength concrete investigated by Klausen [12] had
mean value C200 with fibres, 10 Hz [2]
a 28-day cube strength of fcm = 40MPa. The approach by
Hsu [17] which according to [18] was developed based on
Fig. 9. Comparison between Woehler curves and our experimental test 783 tests conducted on normal-strength concrete between
results 1943 and 1981, should be highlighted at this point for its
extensive database. Fig. 10 shows that test results and fa-
tigue models documented in the literature are quite close
though the test frequencies fP = 0.1Hz 1Hz were very to one another. Furthermore, it is obvious that the
low, which for the maximum stress levels tested generally Woehler curve of [1] underestimates these results.
leads to lower numbers of cycles to failure [5]. Fig. 9 and Fig. 10 both demonstrate that the Woehler
Fig. 9 also shows that the numbers of cycles to failure curves of the CEB-FIP Model Code [1] are quite conserva-
for high-strength and ultra-high-strength concretes with fi- tive. In contrast, the new material model approximates the
bres tend to be lower than those for the same types of con- fatigue behaviour of normal-strength, high-strength and ul-
crete without fibres. The mean value of numbers of cycles tra-high-strength concrete much better. The heat treat-
to failure of the ultra-high-strength concrete without heat ment and the use of steel fibres in the experimental inves-
treatment is similar to the test results for high-strength tigations conducted are considered within this model.
concrete. Consequently, the development of the new material model
Fig. 10 shows a comparison between the new materi- based on these experimental investigations is justifiable
al model, the model according to CEB-FIP Model Code [1] [2]. Furthermore, it is obvious that the fatigue strengths of
and the results from single-level fatigue tests and fatigue all the different types of concrete are quite close to one an-
models documented in the literature. By way of an exam- other. The requirement for a strength-dependent reduc-
ple, the comparison is made for a minimum compressive tion in the fatigue reference strength cannot be derived
stress level of Sc,min = 0.05. The experimental results of from this.
the University of Aalborg originate from a pre-published The new Woehler curves and those of the CEB-FIP
research report which is documented in [15]. The mate- Model Code 90 [1] are compared in Fig. 11. The new
rial investigated has a 28-day compressive strength of curves for high maximum compressive stress levels lead to
fcm = 140MPa. higher numbers of cycles to failure (resistance) than the
Fatigue investigations at the University of Kassel current standard ones. However, lower numbers of cycles
were carried out on cylindrical specimens made from ul- to failure are reached for low maximum compressive stress
tra-high-strength concrete with a compressive strength of levels in the range logN > 8, depending on the minimum
about fcm = 160MPa after storage in water [16]. The mini- compressive stress level. Generally, there is a lack of ex-
mum compressive stress levels in these tests were varied perimental test results in this range. The approximation
between 0.06 und 0.075, which resulted in slightly higher was carried out on the safe side for this reason. For practi-
numbers of cycles to failure in comparison with the mini- cal applications this range of stress levels has only limited
mum compressive stress level of Sc,min = 0.05. From a prac- effects on the sum of damage, according to Palmgren-
tical point of view, this deviation is not very relevant. The Miner, because of the very high numbers of cycles to

188 Structural Concrete 13 (2012), No. 3


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

design value of fatigue reference strength fcd,fat [MP a]


1.0 120
Model Code 90 [1]
0.9 New approach Hannover
maximum stress level S c,max [-]

fat = 1.0
0.8 S c,min = 0.8 100
fat = 0.7
0.7
S c,min = 0.6 Additional safety for fatigue
0.6 80 loading, new approach

0.5 S c,min = 0.4 Additional safety for fatigue


loading, Model Code

0.4 S c,min = 0.2 60


0.3
0.2 S c,min = 0.0 40
Woehler curves Model Code 90 [1]
0.1
New Woehler curves Hannover
0.0 20
S ame safety level for fatigue loading
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 CE B-FIP Model Code 90: C70/C80
number of cycles to failure log N [-] New approach Hannover: C120
0
Fig. 11. Comparison between the new Woehler curves and those of CEB- 0 20 40 60 80 100 120 140 160 180 200
FIP Model Code 90 [1] characteristic static compressive strength fck [MP a]

Fig. 12. Comparison between the new approach to fcd,fat and the existing
approach of CEB-FIP Model Code 90 [1]
failure. Fig. 11 demonstrates that the continuous mathe-
matical description of the fatigue behaviour is another ad-
vantage of the new Woehler curves. As presented in sec- Fig. 12 shows the correlations between the compres-
tion 3.2, only two equations are necessary for their sive strength and the design fatigue reference strength ac-
mathematical description. Additionally, there are no dis- cording to the new approach and [1]. The correlations for
continuities at the transition points. constant values of fat = 1.0, which means no additional
reduction (fcd,fat = fcd), and fat = 0.7 are also analysed. It is
4 Integrating the material model into the existing obvious that applying the fatigue reference strength fcd,fat
design concept of the Model Code according to CEB-FIP Model Code 90 [1] to concretes
4.1 Design fatigue reference strength with compressive strengths of fck > 125MPa would lead to
a regressive curve (see Fig. 12).
The new material model should be integrated into the ex- It is absolutely vital to modify the design value fcd,fat
isting design concept of the Model Code [1] as follows so for the fatigue design of concretes with compressive
that easy practical application is assured. strengths higher than fck = 125MPa. Thus, a new design
According to [1], the design fatigue reference value for the fatigue reference strength was determined
strength fcd,fat is determined based on the static compres- based on the investigations so that it can also be applied
sive design strength fcd, which is reduced by a strength- to concretes with compressive strengths of up to
dependent term, here referred to as fat: fck = 200MPa. The required safety level has to be ensured
here. Although the need for a strength-dependent reduc-
f f tion in the fatigue reference strength cannot be derived
fcd,fat = 0.85 cc(t) ck 1 ck
c 25fck0 from Fig. 9 and Fig. 10, this reduction should be included
on the safe side, but to a lesser extent.
fcd fat
Considering the investigations on ultra-high-strength
concrete (Fig. 1), the extrapolation of the regression lines
in accordance with [1], section 6.7.3 (31) for a minimum stress level of Sc,min = 0.05 up to N = 108 load
cycles leads to a corresponding compressive stress level of
where fck0 = 10MPa (reference strength). Sc,max 0.5, which accounts for a loss of 50 % in strength or
bearing capacity of a C200 strength class concrete. In prac-
The derivation of the partial safety factor, c,fat, for ultra- tical terms this represents an extreme loading condition
high-strength concrete is documented in [2]. Hereafter as since such high amplitudes rarely occur in combination
for normal-strength and high-strength concrete c,fat is with low minimum stress levels. It was for this reason that
assigned a value of c,fat = c = 1.5. That value should be the term fat was established in a way that the design fatigue
verified by further investigations. Generally, the additional reference strength fcd,fat for a C200 class concrete amounts
reduction due to the term fat, taking account of fatigue to 50 % of the design value fcd. The new design fatigue ref-
loading, increases as the static compressive concrete erence strength is calculated using Eq. (32):
strength fck rises. However, for concretes of strength
f f
grades above C80, this reduction is even higher than that fcd,fat,Hannover = 0.85 cc(t) ck 1 ck (32)
due to the partial safety factor c,fat = 1.5. For a concrete c,fat 40fck0

with a static compressive strength fck = 120MPa, the cor- fcd fat
responding reduction term is fat = 0.52. This means that
the design fatigue reference strength fcd,fat only amounts where fck0 = 10MPa (reference strength).
to 52 % of the static strength design value fcd. c,fat = c = 1.5

Structural Concrete 13 (2012), No. 3 189


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

The curve of the modified design value is also shown in where


Fig. 12. It is obvious that the function applies up to a com- Scd,min = Sd c,min c/fcd,fat,Hannover
pressive strength of fck = 200 MPa. Furthermore, fatigue Scd,max = Sd c,max c/fcd,fat,Hannover
reference strength values higher than those according to Sd = 1.1 or 1.0 in accordance with [1], section 1.6.4.4
[1] are permitted. Applying the modified fatigue reference c in accordance with [1], Eq. (6.7-2)
strength leads to fat = 0.70 for a C120 concrete. The same
safety level is thus assured for a C120 concrete as was 0.45 + 1.8 Scd,min
Y=
specified for a concrete in the range between C70 and C80 1 + 1.8 Scd,min 0.3 Scd,min
2
according to the CEB-FIP Model Code 90 [1] in the early
1990s. Meanwhile, the extensive scientific research avail-
fck f
able regarding high-strength concrete (see section 3.4) jus- fcd,fat,Hannover = 0.85 cc(t) 1 ck
c,fat 40fck0
tifies this adjustment to the safety level. Thus, the required
safety level is ensured for all concrete grades up to
fck = 200 MPa. At the same time, economical fatigue Level 3:
design of concrete with characteristic static strengths up If the level 2 requirement is not met, the whole spectrum
to fck = 200MPa is possible. of load levels has to be known. As explained in CEB-FIP
Model Code 90 [1], verifying the fatigue requirements in
4.2 Design concept accordance with the PalmgrenMiner summation [19, 20]
is recommended. The limiting fatigue damage has to be
The material model and the proposed design value met.
for the fatigue reference strength will now be integrated
into the design concept of CEB-FIP Model Code 90 j
n
[1]. D= NSiRi Dlim = 1 in accordance with[1], section 6.7.5
i=1 (37)
Level 1:
Detailed fatigue design is not necessary provided the fol- The numbers of resisting cycles to failure NRi are calculat-
lowing requirement for concrete under compressive fa- ed using Eqs. (35) and (36).
tigue loading is met: The new design model comprising the new Woehler
curves and the modified value for fatigue is included in
Sd c,max c 0.45 fcd,fat, Hannover (33) the new fib Model Code 2010 [4].

fck f 4.3 Verification of the design model developed


where fcd,fat,Hannover = 0.85 cc(t 0 ) 1 ck
c,fat 40fck0 with respect to safety requirements

fck0 = 10MPa (reference strength) The evaluation of the concrete fatigue design with regard
to safety requires a joint analysis of Woehler curves and
In such cases, a number of cycles to failure of N 108 is fatigue reference strength. The fatigue reference strength
reached for a minimum stress level of Scd,min = 0. has a dominant effect with regard to the level of safety.
Generally, the level of safety required is ensured by the de-
Level 2: finition of the fatigue reference strength, or rather by the
If the level 1 requirement is not met, the verification refers amount of the reduction in the characteristic static
to a single load level including the dominant fatigue load- strength. The fatigue reference strength in the design con-
ing. The requirements are fulfilled if the number of load cept is used as the reference value for Scd,max and Scd,min.
cycles occurring n is equal to or less than the number of The correlation between the design stresses cd,max or
resisting cycles N: cd,min and the numbers of cycles to failure N is changed
by modifying the fatigue reference strength. For a chosen
nN (34) value of log N, the design stresses decrease with decreas-
ing fatigue reference strength, when the stress levels are
The number of resisting cycles N is calculated for kept constant. At the same time, the correlation between
0 Scd,min 0.80 using the following equations. In the Scd,max or Scd,min (or rather cd,max or cd,min) and N is non-
case of Scd,min > 0.80, the minimum stress level design val- linear and, as a consequence, small changes in fcd,fat lead
ue is taken as Scd,min = 0.80. to disproportionate changes in numbers of cycles to fail-
ure. Additionally, the influence of the fatigue reference
For logN 8: strength increases with increasing concrete strength
8 grades. As an example, Fig. 13 shows the effect of different
log N = (Scd,max 1)
Y 1( ) (35)
reference values for concrete grades C30 and C120. For
this purpose, the characteristic static strength fck, the stat-
For logN > 8: ic design strength fcd and the design fatigue reference
8 ln(10) S Scd,min strength fcd,fat,Hannover are used as reference values. The
log N = 8 + (Y Scd,min) log cd,max
(Y 1 ) Y Scd,min
new Woehler curve is used here. In addition, the dotted
curve is calculated using the Woehler curve and the fa-
(36) tigue reference strength fcd,fat of CEB-FIP Model Code 90

190 Structural Concrete 13 (2012), No. 3


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

maximum compressive stress cd,max [MP a]


ever, subsequent concrete hardening, which is not consid-
S c,min = 0.05
ered in the design concept, leads to decreasing stress lev-
els during the concretes lifetime and thus to a better fa-
tigue capacity. That means additional safety. Two further
positive influences regarding fatigue strength are the redis-
tribution of stresses in damaged zones and the combined
bearing capacity of reinforced concrete. On the contrary,
there are influences such as preliminary damage, over-
loads and the combination of different load levels that
might reduce the calculated fatigue strength of structures.
As yet, the effect of redistribution, preliminary damage
and overloads can hardly be quantified in a way that is
generally valid for all concrete structures.
Altogether, the new approach does not lead to signif-
number of cycles to failure log N [-] icant changes in the fatigue design results for normal-
C120: fcd,fat,Hannover and Woehler curve Hannover strength concrete. For higher-strength concrete, there is
C120: fcd,fat and Woehler curve Model Code 90 still a huge reduction in numbers of cycles to failure. Tak-
C120: fcd, Woehler curve Hannover ing this into account, and considering the hidden safety
C120: fck, Woehler curve Hannover factors described previously, it is ensured that the new de-
C30: fcd,fat,Hannover and Woehler curve Hannover sign approach complies with the level of safety required.
C30: fcd,fat and Woehler curve Model Code 90
C30: fcd, Woehler curve Hannover
5 Conclusions
C30: fck, Woehler curve Hannover
The design concept according to CEB-FIP Model-Code 90
[1] is limited to types of concrete with compressive
strengths up to fck = 80MPa. Furthermore, the fatigue de-
Fig. 13. Analysis of the effect of reference values fck, fcd, fcd,fat and sign concept includes a reduction in the fatigue reference
fcd,fat,Hannover (C30 and C120)
strength, which for high-strength concrete leads to uneco-
nomical design at the ultimate fatigue limit state. This pa-
per presents the results of experimental and theoretical in-
[1]. For simplicity, the minimum stress level Sc,min = 0.05 is vestigations of the compressive fatigue behaviour of
kept constant. high-performance concrete with concrete strengths up to
It is obvious in Fig. 13 that the effect of the new de- fck = 200MPa. A material model for concrete under uniax-
sign model is greater for high-strength concrete than for ial compressive fatigue loading was developed based on
lower concrete grades because of the modified fatigue ref- these extensive experimental investigations. The new ma-
erence strength (see also Fig. 12). For a normal-strength terial model approximates the fatigue behaviour of nor-
concrete C30, both approaches lead to quite similar maxi- mal-strength, high-strength and ultra-high-strength con-
mum compressive stresses cd,max (fcd,fat or fcd,fat,Hannover as crete. Another advantage is the continuous mathematical
reference strength). Additionally, it can be seen that the description of the Woehler curves, which ensures easier
new approach permits higher numbers of cycles to failure practical application. The new design model was devel-
than [1] for the same maximum compressive stress, espe- oped considering the design concept according to CEB-
cially for C120. At the same time, the differences in load FIP Model Code 90 [1]. The required partial safety factor,
cycles between the new and the current standard curves c,fat = 1.5, was confirmed to be applicable for ultra-high-
are significantly smaller than the total difference when fck strength concrete, too. The characteristic compressive ref-
is used as the reference strength. Comparing the curves erence strength fck,fat and the design value of the fatigue
for reference values fcd,fat,Hannover and fck, it is obvious that reference strength fcd,fat were modified. The modified fa-
a huge reduction in numbers of cycles to failure still exists. tigue reference strength ensures the same level of safety
This reduction represents the degree of safety ensured. for a C120 as for a concrete in the range of C70/C80 ac-
The effect of the fatigue reference strength decreases for cording to [1]. The analysis of the assured safety shows
higher numbers of cycles to failure. that the new design model leads to higher numbers of re-
Further aspects have to be considered when evaluat- sisting cycles to failure, especially for high-strength con-
ing concrete fatigue design with regard to safety. General- crete. These improvements are absolutely crucial for prac-
ly, and differently from both design approaches, the deter- tical applications taking into consideration the increase in
mination of the stress levels in experimental tests refers to fatigue-relevant loads. This modification is justified ac-
mean values of the static strength. This means that even cording to the experimental test results currently available
without any safety factors and additional reduction terms for high-strength concrete. At the same time, and most im-
taking account of fatigue loading additional safety is portantly, it was demonstrated that the level of safety re-
generally included in the design concept since fck refers to quired is still assured. Altogether, the new design model
the 5 % quantile and not to the mean value of static leads to safe and economical design. Furthermore, it en-
strength. This additional reduction in number of cycles to ables the prospect of extending the application to ultra-
failure is not included in Fig. 13. Another aspect is that the high-strength concrete with compressive strengths up to
fatigue reference strength refers to 28-day strengths. How- fck = 200MPa. The design model presented is included in

Structural Concrete 13 (2012), No. 3 191


L. Lohaus/N. Oneschkow/M. Wefer Design model for the fatigue behaviour of normal-strength, high-strength and ultra-high-strength concrete

the new fib Model Code 2010 [4], although the concrete 13. Grnberg, J., Oneschkow, N.: Grndung von Offshore-
grades are limited to C120. Windenergieanlagen aus filigranen Betonkonstruktionen
unter besonderer Beachtung des Ermdungsverhaltens von
hochfestem Beton. Final report for BMU joint research pro-
Acknowledgements
ject, Leibniz University of Hannover, 2011.
14. Anders, S., Lohaus, L.: Polymer- und fasermodifizierte
This research project was supported by the German Re- Hochleistungsbetone fr hochdynamisch beanspruchte Ver-
search Foundation (DFG) within the scope of priority pro- bindungen wie Grouted Joints bei Windenergieanlagen.
gramme 1182 Sustainable Building with Ultra-High-Per- Final report for research scholarship T 4/2002, Stiftung In-
formance Concrete (UHPC). The authors would like to dustrieforschung, Hannover, 2007.
express their gratitude for the financial support. 15. Lohaus, L., Anders, S.: High-cycle Fatigue of Ultra-high Per-
formance Concrete Fatigue Strength and Damage Devel-
References opment. Fdration Internationale du Bton, Proceedings of
2nd International Congress, Naples, 2006.
1. CEB Comit Euro-international du Bton: CEB-FIP Mod- 16. Fehling, E., Schmidt, M., Teichmann, T., Bunje, K., Borne-
el Code 90. Bulletin dInformation, No. 213/214, Thomas mann, R., Middendorf, M.: Entwicklung, Dauerhaftigkeit
Telford Ltd., London, 1993. und Berechnung Ultra-Hochfester Betone (UHPC). Schrif-
2. Wefer, M.: Materialverhalten und Bemessungswerte von tenreihe Baustoffe und Massivbau, University of Kassel, No.
ultrahochfestem Beton unter einaxialer Ermdungsbean- 1, 2004.
spruchung. Dissertation, Leibniz University of Hannover, In- 17. Hsu, T.: Fatigue of Plain Concrete. In: ACI Journal, JulAug
stitute of Building Materials Science, 2010. 1981, pp. 292305.
3. Lohaus, L., Wefer, M., Oneschkow, N.: Ermdungsbemes- 18. Pfanner, D.: Zur Degradation von Stahlbetonbauteilen unter
sungsmodell fr normal-, hoch- und ultra-hochfeste Betone. Ermdungsbeanspruchung. Research reports, VDI Reihe 4,
In: Beton- und Stahlbetonbau, vol. 106, No. 12, 2011, pp. No. 189. VDI Verlag, Dsseldorf, 2003.
836846. 19. Palmgren, A.: Die Lebensdauer von Kugellagern. Zeitschrift
4. fib International Federation for Structural Concrete: Mod- VDI, 68, 1924, pp. 339341.
el Code 2010, final draft, Volume 1 and 2, March 2012. 20. Miner, A. M.: Cumulative Damage in Fatigue. Journal of
5. Petkovic, G., Stemland, H., Rosseland, S.: High Strength Applied Mechanics, No. 12, 1945.
Concrete SP3 Fatigue, Report 3.2 Fatigue of High Strength
Concrete. SINTEF Structures and Concrete, Trondheim,
Aug 1992.
6. Petkovic, G., Lenschow, R., Stemland, H., Rosseland, S.: Fa-
tigue of High-Strength Concrete. In: High-strength concrete:
Second International Symposium, American Concrete Insti-
tute (ACI Special Publication 121), Detroit, 1990, pp.
505525.
7. Grnberg, J., Lohaus, L., Ertel, C., Wefer, M.: Mehraxiales
mechanisches Ermdungsmodell von Ultra-Hochfestem Be-
ton (Experimentelle und analytische Untersuchungen). In:
Beton- und Stahlbetonbau, vol. 102, No. 6, 2007, pp. 388 Prof. Dr.-Ing. Ludger Lohaus Dipl.-Ing. Nadja Oneschkow
398. lohaus@baustoff.uni-hannover.de n.oneschkow@baustoff.uni-hannover.de
8. Bennett, E., Muir, S.: Some Fatigue tests of high-strength Leibniz Universitt Hannover
concrete in axial compression. In: Magazine of Concrete Re- Institute of Building Materials Science
search, vol. 19, No. 59, June 1967. Appelstrae 9A
9. Hohberg, R.: Zum Ermdungsverhalten von Beton. Disserta- 30167 Hannover
tion, TU Berlin, 2004.
10. Weigler, H., Freitag, W.: Dauerschwell- und Betriebsfestigkeit
von Konstruktionsleichtbeton. Deutscher Ausschuss fr
Stahlbeton, No. 247, 1975.
11. Wilrich, P.-T., Henning, H.-J.: Formeln und Tabellen der an- Dr.-Ing. Maik Wefer
gewandten mathematischen Statistik. Springer Verlag, Ber- Leibniz Universitt Hannover
lin, 2003. ForWind Center for Wind Energy Research
12. Klausen, D.: Festigkeit und Schdigung von Beton bei hu- Appelstrae 9A
fig wiederholter Beanspruchung. Dissertation, TU Darm- 30167 Hannover
stadt, 1978. maik.wefer@forwind.uni-hannover.de

192 Structural Concrete 13 (2012), No. 3