View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Infoscience - École polytechnique fédérale de Lausanne
Engineering Frorrure Mechanics Vol. 35. No. l/2/3. pp. 565-571, 1990 0013-794419053.00+ 0.00
Printed in Great Britain. Pergamon Press pk.

FAILURE OF DAM CONCRETE SUBJECTED TO SEISMIC

LOADING CONDITIONS
E. BROHWILER and F. H. WITTMANN
Swiss Federal Institute of Technology. Laboratory for Building Materials, Lausannc, Switzerland

AMtract-The numerical simulation of concrete dams subjected to earthquake loading requires

realistic material laws which take into account seismic loading conditions. Dynamic tests were
performed in order to examine the effect of initially applied compressive loading on material
properties of a dam concrete at high tensile deformation rates. The test results show that no
important fracture property is reduced at high deformation rates; the tensile strength and the
specific fracture energy GF show a high rate sensitivity. However, dynamic compressive pre-loading
leads to a reduction of the fracture properties at both quasi-static and high deformation rates.

1. INTRODUCTION
THE NUMERICAL simulation of concrete dams subjected to earthquakes includes not only the
structure and its material properties, but also the effect of soil-structure and reservoir-structure
interaction. In view of this very complex analysis, simple but realistic material laws which take into
consideration seismic loading conditions, are required for the structural analysis.
Seismic loading is characterized by mainly horizontal motions. Irregular vibrations of a
frequency of 1-15 Hz and ground acceleration of at most 0.2-0.5 g occur. The structure and hence
the material is exposed to alternating excitations at high loading rates, i.e. strain rates C of
10m4< 1 < 1 per s. During an earthquake, several parts of the dam may suffer tensile loading with
subsequent crack formation. Thus, the safety of these large unreinforced concrete structures when
subjected to earthquakes is often controlled by the tensile behaviour and cracking of concrete.
Crack formation in dams and the consequences may be analysed by means of fracture mechanics
methods.
In the last few years, important efforts have been undertaken in order to study, both
theoretically and experimentally the behaviour of concrete under high loading rates[l]. However,
these studies have not examined the material properties at high loading rates following dynamic
compressive loading. Before concrete is loaded, microcracks already in both the cement matrix and
along aggregate inclusions, may grow under compressive loading and consequently reduce the
material properties in tension. Therefore, for the current investigation dynamic tests were carried
out in order to study the effect of initially applied compressive loading on the fracture properties
of dam concrete at high deformation rates[2].

2. FRACTURE PROPERTIES OF DAM CONCRETE AT

QUASISTATIC LOADING RATES
Concrete used for the construction of dams and common concrete differ mainly in the grain
size and in the nature of the aggregates. Because of their origin in the neighbourhood of the dam,
the crushed aggregates of dam concrete may have poorer strength than the rolled grave1 used for
common concrete, and, consequently, strength of dam concrete may be lower. Therefore, an
“extrapolation” from common concrete to dam concrete cannot be made directly.
In order to determine the specific fracture energy G, of dam concrete, wedge splitting tests
were performed. The principle of this test method[3] consists in splitting a notched specimen by
means of wedges pressed between rollers placed on the top of the specimen. Wedge splitting
specimens have a large fracture area compared to the specimen weight. Drilled cores can easily be
used as specimens. Therefore, this method is suitable for testing dam concrete. Cylindrical wedge
splitting specimens drilled from three existing dams were tested[2], and bilinear strain-softening
relations were evaluated using a numerical method[4].
565
566 E. BRUHWILER and F. H. WITTMANN

G,-values of 175,235 and 257 N/m were obtained for these three different dam concretes. This
means that the Gr-value of dam concrete is two to three times larger than the Gr of common
concrete, Since the tensile strength and the modulus of elasticity of dam concrete are smaller than
those obtained for common concrete, the characteristic length lc,,= E. GF/fi of dam concrete is up
to 10 times larger than for common concrete; &-values of 1100, 1300 and 1700 mm were obtained
for the three dam concretes tested. The evaluated strain softening diagrams reveal very large crack
width values; maximum crack width of 0.40 mm was evaluated for one dam concrete. It can be
concluded that nonlinear models like fracture process zone models should be applied even for dam
structures like arch dams, buttress dams and the crest part of gravity dams[2].
It is interesting to note that the fracture surface of dam concrete specimens was characterized
by mainly aggregate failures. This may be explained by the relatively low strength of the aggregates
used for the construction of dams. However, it seems that a lot of energy is need to break aggregates
like gneiss which has a laminar texture. These findings indicate that the high Gr-value of dam
concrete is rather the result of the nature and material properties of the aggregates than of the
aggregate size.

3. DYNAMIC TESTS
The complete tensile stress-strain diagram of concrete, which can be subdivided into two
portions, is investigated by means of two types of tests:

-uniaxial tests to study the ascending branch of the stress-strain curve; and
-wedge splitting tests to examine the decending branch with strain-softening.

3.1. Test specimens

In both series, cylindrical specimens having a diameter of 200 mm were tested. These concrete
cylinders were drilled from blocks which have been poured on the structural site of a dam in the
Swiss Alps. This dam concrete with a cement content of 250 kg/m3 and a water-cement ratio of
0.5, was made of crushed aggregates with a maximum grain size of 80 mm. These aggregates consist
of metamorphic rocks (granite, gneiss, mica shist, quartz) and limestone.
The strength of the dam concrete was determined on cylinders with a diameter of 200 mm and
a height of 600 mm. The following values were obtained for the concrete at an age of two-years:
compressive strength f, = 39 N/mm’, tensile strength f, = 2.4 N/mm*, modulus of elasticity
E = 36400 N/mm2 (both in tension and in compression).
The aim of the dynamic tests is to seize the effect of strain rate and compressive preloading
on the fracture properties; the absolute test values are only of secondary importance. Therefore,
test results are normalized with respect to the quasi-static properties of the unpreloaded dam
concrete. Under the assumption that the observed effects of rate and preloading are also valid for
other dam concretes, this normalized presentation enables a more general use of the test results
(Section 4).

3.2. Tensile tests

A detailed description of these experiments is given in[5]. Uniaxial tests on cylindrical dam
concrete specimens (Fig. la) were controlled by means of deformation (strain). In order to examine
the effect of preceding dynamic compressive loading on the tensil,e behaviour at high strain rates,
the following “loading histories”, representing typical seismic loading conditions, were chosen
(Fig. lb):

A. Pure tension: tensile loading to failure at different strain rates.

B. Tension after simple compression: compressive loading to a certain strain level and
subsequent tensile loading to failure at different strain rates.
C. Tension after several cycles of compression: repeated compressive loading at constant
amplitude and a certain mean strain level with subsequent tensile loading to failure at different
strain rates.
 Failure of dam concrete 567

(a) (b)

A. Pure tension
t
l6L,
B. Tension after simple compression

l
6,

L+-
i c?kl I
300 600 B -0.75
l#J=200 l t

+%
II
C. Tension after several cycles of compression

6
AC er I/T N

:: q T2Vl P$‘ooo CHzl

Cl
‘O
:23 0.24 -0.48
-0:24 IO x
::
1 cc

-----------
Strain rates

Fig. 1. Tensile tests. (a) Dimensions (in mm) of the specimens. (b) Schematic presentation of the “loading
histories” and the chosen parameters.

The tensile strain rate i of the final tensile loading was varied between low5 and lo-’ per s.
The strain rate of i = 10e5 per s was considered to be the quasi-static strain rate. In a deformation
controlled tensile test, an overall load-deformation-curve with a descending branch can generally
be recorded. This complete curve was measured at low strain rates. However, at high strain
rates, the tests could no longer be controlled and subsequently the specimens failed in a violent
manner.
For each series, a power law relating the tensile strength f, and the modulus of elasticity E
to the tensile strain rate has been determined by means of a least squares fit. With the help of a
stochastic approach Mihashi and Wittmann[6] have predicted a power law for the rate-sensitivity
of strength. The evaluated power functions are shown in Fig. 2. All test results were related to the
quasi-static tests without preloading (series A). In order to compare the “loading histories”, the
“intensity” of the preceding compressive loading was quantified in two ways [5]: first, the energy
dissipated during compressive preloading was determined from the area under the compressive
stress-strain diagram (including the energy dissipated during the load-cycles). “Loading history”
C3 showed by far the highest energy dissipation. The compressive strain level is another value

1.6 I- (a)
,A (b)

1.6

A
,c2
,c3 ‘Cl
‘B

Fig. 2. Effect of compressive preloading and high strain rates on: (a) the tensile strength; and (b) the
modulus of elasticity.
568 E. BROHWILER and F. H. WI’ITMANN

which allows us to quantity the “intensity” of the compressive preloading. The highest compressive
strain level was applied in series B.
The results of pure tension tests (series A) are compared to those of tensile tests after initially
applied compressive loading (series: B, Cl-C3). The main findings are:

-Irrespective of preloading, higher strain rates yield higher tensile strength values (Fig. 2a).
However, this rate sensitivity is reduced with increasing intensity of compressive preloading. At
quasi-static strain rates, the tensile strength is the same for both preloaded and unpreloaded
concrete. This means that only the rate sensitivity is reduced due to compressive preloading.
-The exponent p = 0.081 in the power law for the tensile strength of the tested dam concrete
is larger than for common concrete, for which p-values of about 0.055 are found[l]. Rate sensitivity
depends on the concrete quality; i.e. the lower strength concrete (dam concrete) is more rate
sensitive than the higher strength concrete (common concrete).
-The modulus of elasticity of unpreloaded dam concrete increases with higher strain rates
(Fig. 2b), whereas the modulus of elasticity of preloaded concrete is independent of strain rates.
The higher the strain level of the compressive preloading, the stronger the reduction of the modulus
of elasticity with respect to unpreloaded concrete.
-No rate sensitivity could be observed for the failure strain; i.e. tensile strain at maximum
tensile stress. The failure strain increases with growing intensity of compressive preloading (and
thus decreasing modulus of elasticity), and the energy stored in the specimen at maximum stress
is the same irrespective of the “loading history”.
-Examination of fracture surfaces of unpreloaded concrete revealed an increase of the total
area of broken aggregates with higher rates. Under the influence of compressive preloading, the
amount of aggregate failure was the same for all strain rates tested.

3.3. Wedge splitting tests

The wedge splitting tests on the same dam concrete as used for the tensile tests can be
subdivided into two series:
-The influence of COD-rate on the specific fracture energy Gr and the strain softening
diagram has been studied using cylindrical specimens shown in Fig. 3(a). The main drawback of
this specimen shape is that it needs a longitudinal reinforcement in order to prevent shear failure
of the cantilevers. For the highest COD-rate of about 10,000 times the quasi-static rate, a strain
rate i of 5 x 10m2per s at the notch tip was found by a rough estimation. (It should be noted that
this strain rate descreases with propagating fracture zone.)
-Fragments of cylinders previously tested and preloaded in the tensile test series, were used
to investigate the effect of compressive preloading on Gr and the strain-softening diagram at
quasi-static COD-rates (Fig. 3b). As in the uniaxial tensile test, the fracture surface of these wedge
splitting specimens is perpendicular to the axis of the cyclinder.

(al

1
@I 210
300

Fig. 3. Wedge splitting tests: Dimensions (in mm) of the specimens used in order to investigate the effect
of: (a) crack opening displacement rate on G,; and (b) compressive preloading on G, at quasi-static
COD-rate.
 Failure of dam concrete 569

(b)

0 0.2 0.4 0.6 0.6 I.0 1.2 1.4 1.6 1.6 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

COD(mm) Crack width w (mm)

Fig. 4. (a) Splitting force-crack opening displacement-mean curves at four different relative COD-rates
(t$ = 1: quasi-static COD-rate). (b) Bilinear strain-softening diagrams.

Remark. The specimen dimensions are relatively small compared to the aggregate size, i.e.
smaller than three times the maximum grain size. Nevertheless the test results showed relatively
low scatter. This may be explained by the observed fracture behaviour of dam concrete: The crack
does not travel around the aggregates; it goes straight through the aggregates.
The splitting force F,-COD-curves from both tests series are given in Figs 4(a) and 5(a). The
relation between the COD-rate and GF is also expressed by a power law. The numerically evaluated,
bilinear strain softening relations are presented in Figs 4(b) and 5(b). The main results are listed
below.
Eflect of COD-rate.
-The splitting force F,-COD-curves at different COD-rates have similar shapes (Fig. 4a).
Higher COD-rates yield higher G,-values and higher maximum splitting forces values. The G,-value
at a COD-rate of lo4 times the quasi-static rate is some 80% higher than the quasi-static G,-value.
-The bilinear strain softening diagrams show that the values of the maximum crack width
are the same irrespective of the COD-rate. The rate sensitivity of the strain softening diagram and
thus of the GF-value can be attributed to the rate sensitivity of the tensile strength (Fig. 4b).
Eflect of compressive preloading.
-The specific fracture energy GF at quasi-static deformation rates decreases with increasing
intensity of initially applied compressive loading.
-The maximum splitting force is not affected by compressive preloads. (This result coincides
with findings from the uniaxial tests.) The slope of the descending branch of the F,-COD-curve
becomes steeper with increasing intensity of compressive preloading (Fig. 5a).
-With increasing intensity of compressive preloading, the bilinear strain softening diagrams
show smaller maximum crack width values (Fig. 5b).

4. APPLICATION OF THE TEST RESULTS

With the results of our dynamic tests on dam concrete under seismic loading conditions,
material models describing tensile behaviour (including fracture) under seismic loading conditions
may be developed. The test results can be summarized as follows:

-No important fracture mechanics property is reduced with higher deformation rate. The
rate-sensitivity of certain properties is small and may be neglected in order to simplify matters.
-Dynamic compressive loading leads to material damage, which results in a decrease of the
values of the fracture properties. With increasing “intensity” of initially applied compressive
loading, this reduction is more pronounced.
570 E. BROHWILER and F. H. WIlTMANN

-1 A (a)

E
t
c
6

I I I I I

0 0. I 0.2 03 0.4 0.5

COO (mm) Crack width w (mm)

Fig. 5. Effect of compressive preloading on Gr and strain softening at quasi-static COD-rate. (a) Splitting
forcexrack opening displacement-mean curves. (b) Bilinear strain-softening diagrams.

-The relationship between mechanical properties and deformation rate can often be described
by a power law; i.e. on a doubly-logarithmic scale, strength increases linearly with rate. As a
consequence, an exact knowledge of the rate of deformation is not always necessary.

On the basis of these general findings a procedure for the development of material laws is
proposed:

-The tensile behaviour of concrete is subdivided into the stage preceding the reaching of
tensile strength and a post-peak stage with strain softening.
-A normalized material law is introduced, where stresses and deformations are related to a
characteristic value such as the tensile strength, failure strain, maximum crack width or the ratio
G/1; (Fig. 5).
-The absolute material law used for the analysis is obtained by a transformation which is
based on the knowledge of the quasi-static properties S, of unpreloaded concrete.
-Simple relations for the absolute material properties S as a function of the seismic “loading
history” (i.e. deformation rate d and preloading D) are proposed in[2]; they can be generalized as
follows:
The effect of deformation rate on mechanical properties is expressed by a power law:

s = S,(ti)P (1)
where u is the (constant) deformation rate and p is an exponent determined from test results. The

___-----
effect of compressive proloading on mechanical properties is quantified according to damage
mechanics, e.g.
S=S,(l -II,) (2)

o/f, o/f,

‘6 ‘L E”

s/e.
s/f, -

Cl
c=w It
GF

c2

Fig. 6. Schematic presentation of a normalized material law for concrete in tension.

C
 Failure of dam concrete 571

with D, as a damage scalar (0 < D, < 1)describing the material damage due to compressive
preloading.
-In view of a, general applicability, aspects such as multiaxial dynamic loading and variation
of deformation rates as well as the effect of unloading/reloading cycles must be considered.
A constitutive relation for plane stress problems with particular regard to the effects of strain
rate, progressive damage and strain softening in discrete cracks has recently been proposed for
structures like dams[7].

5. CONCLUSIONS
-The most important difference in the fracture mechanics properties of normal and dam
concrete lies in the specific fracture energy GF. GF of dam concrete can be several times greater
than GF of common concrete. This finding is explained rather by the nature and the fracture
properties of aggregates than by the larger maximum aggregate grain size of dam concrete.
-No important fracture mechanics property is reduced at high loading rates. The tensile
strength and the specific fracture energy show a high rate sensitivity. The rate effect on the G,-value
can be attributed to the rate sensitivity of the tensile strength.
-Dynamic compressive preloading leads to a reduction of fracture mechanics properties.
Especially the specific fracture energy GF of preloaded dam concrete decreases considerably with
increasing “intensity” of the preloading.
-The stress-strain diagram under tensile loading can be subdivided into two portions: an
ascending branch with little non-linearity, and a descending branch with pronounced strain
softening. Diagrams normalized with respect to the quasi-static properties of unpreloaded concrete
clearly reveal the effect of, e.g. strain rate and compressive preloading. This normalized represen-
tation can serve for the elaboration of material laws.

Acknowledgement-This research was supported by the Swiss Federal Office for Dam Safety under the direction of Dr R.
Biedermann.

REFERENCES
[I] S. Mindess and S. P. Shah (Eds), Cement-based composites: strain rate effects of fracture. MRS Symposia-Proc., 64,
Materials Research Society, Pittsburgh (1986).
[2] E. Briihwiler, Fracture machanics of dam concrete subjected to quasi-static and seismic loading conditions. Doctoral
Thesis, Laboratory for Building Materials, Swiss Federal Institutue of Technology Lausanne, (1988) (in German with
an extended summarv in Enalish).
[31 E. Brtihwiler and F. H. Wit&a& The wedge splitting test, a method of performing stable fracture mechanics tests.
Engn Fracture Mech. 35, 117-125 (1990).
[41 P. E. Roelfstra and F. H. Wittmann, Numerical methods to link strain softening with failure of concrete, in Fracture
Toughness and Fracture Energy (Edited by F. H. Wittmann), pp. 163-175. Elsevier, Amsterdam (1986).
PI E. Briihwiler, Experiments on the influence of compressive preloading on the behaviour of dam concrete in tension
at high strain rates. Test Report, Laboratory for Building Materials, Swiss Federal Institute of Technology, Lausanne
(1987) (in German).
El H. Mihashi and F. H. Wittmann, Stochastic approach to study the influence of rate of loading on strength of concrete.
HERON (The Netherlands) 25 (1980).
[71 Ph. Chappuis, Nonlinear modelling of the material behaviour of concrete subjected to dynamic loading. Doctoral
Thesis, Swiss Federal Institute of Technology Zurich, IBK-Report No. 155, Birkhauser Basel- Boston-Stuttgaert (1987)
(in French).

(Received for publication 16 November 1988)