Escolar Documentos
Profissional Documentos
Cultura Documentos
Influence of concrete interface friction on the seismic performance of fractured concrete dams
Bogart C. Mendez U.
Dept. of Geotechnical Engineering, Instituto de Ingeniera, UNAM, Mexico D.F., Mexico
Miguel P. Romo O.
Dept. of Geotechnical Engineering, Instituto de Ingeniera, UNAM, Mexico D.F., Mexico Keywords: sliding friction, concrete dam, hydrodynamic pressure ABSTRACT: Shaking table experiments were carried out on concrete-concrete interfaces to develop a continuous friction model, which was validated by implementing it in the commercial finite difference code FLAC3D to reproduce laboratory measurements of displacements with it. The analyses showed that constant and Coulomb friction underestimate laboratory measurements. Once validated, the friction model was used in a numerical FLAC3D example of a fractured concrete dam to investigate the influence of friction variation on the permanent relative displacements of a hypothetical detached block. The analysis considered hydrodynamic pressures on the upstream face of dam as well as uplift forces at the crack. Results showed that the variable friction model yielded larger displacements as compared to constant and Coulomb friction, leading to conservative (safe) predictions.
1 Introduction
Concrete sliding on concrete is common in a variety of engineering problems. One of the problems that have drawn the attention of many researchers is that of cracked concrete dams. Furthermore, it is generally acknowledged that the construction procedures used in rolled compacted concrete dams, for example, in which the material is placed in somewhat irregular horizontal layers up to 50 cm thick, could contribute to localize and generate possible fault surfaces, thus leading to potential sliding of blocks formed in some part of the dams body when subjected to seismic action. Background exists on this matter in the technical literature for gravity dams (e.g., Pekau and Yuzhu, 2004) and concrete arch dams (e.g., Malla and Wieland, 2006). Published information concerning this problem generally presents analyses performed assuming a single friction coefficient, i.e., dynamic friction coefficient, d, being equal to the static friction coefficient, s. This is a simplification of the actual phenomenon. It is well known that friction during sliding can be considered constant only under constant sliding velocity conditions, which hardly occurs during seismic events. Therefore, a varying friction coefficient should be used for more realistic numerical investigations. Accordingly, an experimental research was carried out in this investigation to study the variation of concrete to concrete friction during shaking. Laboratory tests were performed with a one-axis shaking table on a concreteconcrete interface. The interface studied was obtained by the contact of a concrete block resting on top of another concrete block. The top block was free to slide over the fixed bottom block. The free sliding nature of the experiments intrinsically accounted for the non-linear variation of the friction coefficient and energy transfer through the contact interface along the model (Mndez, 2004). This is an important feature of the experiments because it is known that friction coefficients estimated from sliding induced through forcing apparatus (e.g. pistons, springs, etc.), are likely to depend on the stiffness of the force applying device (Cambou, 1974). From the experimental results a friction model was proposed in terms of dynamic input velocity rate and material experimental constants. Afterwards, the model was implemented in the commercial finite difference code FLAC3D (Itasca, 2005) to perform a numerical simulation of the seismic response of a hypothetical fractured concrete dam to evaluate dam collapse due to potential block sliding and / or overturning. To validate the proposed friction model the experiments were simulated numerically, achieving satisfactory results. Once the model was validated, the fractured dam analysis was performed considering the coupled actions of sliding across the concrete interface, hydrodynamic pressures acting over the upstream-face (considering incompressible reservoir conditions) and uplift forces into the crack. For comparison, the results of an analysis using both Coulomb and constant friction coefficient during sliding are presented. It is shown that these assumptions lead to underestimations of the permanent seismic-induced displacements in detached blocks within the dam body.
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
such that = s
Attached to actuator
Figure 1. Apparatus used for the experiments. Static (a) and dynamic tests (b) configuration During the static tests the shaking table remained still while the inclination of the pad was increased to a value up to the onset of sliding of the top block (see figure 1a). This angle was regarded as the concrete-interface static friction angle: s. The dynamic experiments included monitoring the sliding response of the top block under a harmonic excitation with frequency and amplitude of 2 Hz and 6 cm, respectively. The friction coefficients, both static and dynamic, were computed for undisturbed conditions only (less than 2000 shearing cycles). It is very important to stress this out, because the frictional characteristics of concrete interfaces heavily depend on wear rate, as noted by other related investigations (e.g., Botero et. al., 2007). This wear dependency of concrete friction (state variable) was not considered in this paper.
3 Laboratory results
The acceleration histories recorded during the experiments are presented in this section. These records are periodic as the input signal was harmonic. Therefore, only one acceleration cycle is enough to represent the overall behaviour of the problem being analyzed. In figure 2a, acceleration records of the input signal (bottom block) and sliding block (top block) are presented for a concrete - concrete interface, and figure 2b shows velocity histories obtained by numerical integration of acceleration measurements. Figure 2 clearly shows the difference between stick and slip phases of the model by the coincidence and separation, respectively, of the acceleration and velocity records of the input motion and the top block response. Before point A in figure 2, both the input motion and the top block acceleration histories coincide thus indicating no relative displacement in the sliding interface develops. At point A, the forces at the sliding interface are at limit equilibrium, i.e., the top block is at the onset of sliding, meaning that yield acceleration has been reached (Newmark, 1965). Once yield acceleration has been attained, it is generally assumed that no further acceleration will be transferred through the sliding interface (Newmark, 1965; Yegian and Lahlaf, 1992; Jangid, 2000; Yegian and Kadakal, 2004). However, it is observed in figure 2 that some percentage of the input acceleration keeps transferring across the interface and it becomes constant from point B to point C, where acceleration begins to decrease until coupling is reached at point D. Accelerations more than 100% higher than the yield acceleration were transmitted throughout the interface for this test conditions. This can be seen by comparing the difference in acceleration between points A and B in figure 2a. This behaviour is attributed to the gradual decrease of friction coefficient when sliding starts: the decreasing shear strength at the interface allows that some of the input energy keeps on transferring as the friction coefficient decreases. This transition from stick to slip conditions has been documented by Rabinowicz (1951), theoretically only, and by Chaudhuri and Hutchinson (2005) under static tilt tests and by Mndez (2004) and Mndez and Romo (2006) from shaking table experiments. Note that this does not follow
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
(g)
C D
B C
4.75
4.85
4.95 t (s)
5.05
5.15
4.85
4.95 t (s)
5.05
5.15
(a)
(b)
s = 29.65
(t)
10.00 4.00 4.25 4.50 4.75 5.00 5.25 5.50 5.75 6.00 t (s)
(t ) = S
&& && U g (t ) U (t ) g
(1)
where g(t) is the bottom block acceleration time series, (t) is the top block acceleration time series, g is the gravitational acceleration, s is the static friction coefficient and is the instantaneous friction coefficient at time t. Equation 1 is applied in each time step to laboratory measurements of acceleration, thus obtaining friction time histories like the one depicted in figure 3. By considering the work-energy theorem or the inertia forces at the interface (limit analysis), inertial effects are considered in the friction phenomenon (as shown in equation 1), which is very important for sliding friction variation during dynamic events. This condition is generally not accounted for when dealing with friction-related problems, where it is very common to use Coulombs friction law (e.g., Yegian and Kadakal, 2004; Kafali et. al., 2007), which was derived from constant velocity-rate experiments to obtain dynamic friction coefficients. Therefore, Coulombs law should be used with caution when modelling dynamic problems, because seismic-induced displacements may be underestimated, as will be shown later. It is very interesting to note that laboratory estimates of sliding friction yields values higher than the static one, when the opposite holds in Coulomb-like experiments, i.e., under constant velocity rate. A higher d than s means that a greater force is needed to maintain motion than to initiate it. This is a very important fact because a greater friction coefficient translates into higher energy level crossing through a sliding interface, thus diminishing energy dissipation and consequently rising structural demand (Mndez et. al., 2006; Botero and Romo, 2005).
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
5 Friction modelling
Based on experimental results, which show a gradual decrease of friction coefficient as a function of excitation velocity rate, a non linear exponential model is proposed herein to capture the physics of the problem. Similar models have been used in the past by several authors (e.g., Constantinou et al., 1990; Andreaus and Casini, 2001; Naboulsi and Nicholas, 2003) which consider constant velocity conditions. In this study, it was observed that velocity-changes affect friction variation, rather than constantvelocity conditions as assumed by other researchers. In addition, one should be aware that constant velocity conditions will hardly ever be achieved in a sliding interface during seismic episodes. The proposed friction law is outlined in figure 4a, along with Coulombs friction law for comparison purposes. In this figure is the instantaneous friction coefficient, s and k are Coulombs static and kinetic friction coefficients, respectively,
& & U rel is interface relative velocity, U g is excitation velocity, sgn is the sign function, exp is the base
1
of natural logarithms, and is a material and dynamic response-dependant coefficient obtained from shaking table experiments under desired excitation conditions. The proposed friction law is given by equation 2:
d & = s exp U g dt
(2)
Observe that equation 2 does not produce friction coefficients greater than s, due to its exponential nature. In order to achieve friction coefficients larger than the static value, s was substituted by the maximum friction coefficient estimated from laboratory measurements, max. Figure 4a shows that the exponential law proposed herein describes a smooth transition into body motions. This allows a more realistic modeling of stick-slip behavior, as this friction law is not piecewise defined. Transition ranges from steady to abrupt according to materials in the interface and dynamic conditions. Note that the exponential friction law varies k within a broad range of values according to its dependence on excitation velocity rate, interface materials, dynamic response, and so forth. All this factors are included in the coefficient of the friction law. Figure 4b shows instantaneous friction coefficients computed with the proposed exponential friction law given by equation 2. Note in figure 4b that the exponential friction model satisfactorily matches the overall behavior observed in the experimental results. The coefficient in equation 2 was found to be 1.9 for the test conditions.
& & = S S sgn Urel + K sgn Urel
)
45.00 40.00 35.00 30.00 25.00 20.00 15.00 10.00 4.00 4.25 4.50 4.75 5.00 5.25 5.50 5.75 6.00 s = 29.65 Laboratory Computed
= S exp
d & Ug dt
(t)
t (s)
Coulomb's law Exponential law
t (s)
(a)
(b)
Figure 4: (a) Coulombs friction law compared to the proposed exponential friction law. (b) Measured and computed friction angles using the proposed exponential friction law depicted in figure 4a
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
obtained using constant friction diverged the most from measured displacements, yielding an underestimated value. When using Coulombs law, i.e., a piecewise discontinuous friction variation, (s = 29.65, d = 22.50), the divergence was reduced but the result still underestimated measurements. Finally, when considering a full continuous variation of friction, the result was least divergent from measurements and the computed value was larger than measurements. These results show the importance of considering friction variation for numerical simulations of friction-related problems. It should be mentioned that the friction model proposed herein assumes a constant static friction angle and constant decay rate during sliding ( in equation 2). However, other investigations have shown that concrete to concrete friction is modified significantly as shearing continues for long time (Botero et. al., 2007). This should be kept in mind when performing numerical simulations of long-term seismic behaviour of concrete structures containing cracks or interfaces.
(a)
(b)
Figure 5. (a) FLAC3D numerical model, (b) computed relative displacements for all cases considered
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
0.20 0.30
y
1:0 . 72 5
96.50 m
70.00 m
(a)
(b)
Figure 6. (a) View of dam and foundation considered, (b) details of analyzed dam
3.00 2.00 1.00 0.00 -1.00 -2.00 -3.00 0.00 5.00 10.00 t (s) 15.00 20.00
(a)
(b)
Figure 7. (a) Response spectrum of input signal, (b) Acceleration history at bottom of detached block
66.50 m
103.00 m
1:0.19
36.50 m
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
25 20 15 10 6.00 Constant friction Coulomb friction Variable friction 6.20 6.40 t (s) 6.60 6.80 7.00
Urel (m)
5.00
10.00
15.00
20.00
t (s)
Figure 9. Block displacement relative to dam body
The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG) 1-6 October, 2008 Goa, India
9 Conclusions
The numerical investigation carried out in FLAC3D to simulate a fractured concrete dam considered hydrodynamic pressure, as well as uplift pressure in the crack, and three different cases of interface friction: constant, Coulomb and variable friction. The latter was simulated with a model specifically derived from shaking table experiments on a concreteconcrete interface implemented in FLAC3D through FISH built in programming language. Results showed that constant and Coulomb friction underestimated laboratory measurements and yielded 47% and 39% smaller relative displacements for constant and Coulomb friction with respect to the variable friction case. This confirms that the friction should be considered variable in contact dynamic problems. The friction model advanced in this paper can be improved by taking into account the dependency of static friction angle and decay rate of variable friction with shearing cycling. This should be considered when analyzing longterm behaviour of sliding in concrete. Also, more than one crack with inclinations other than 0 should be taken into account to analyze several potential failure scenarios.
10 References
Andreaus U. and Casini P. (2001). Dynamics of friction oscillators excited by a moving base and/or driving force, Journal of Sound and Vibration, 245(4), 685-699 Botero E. and Romo M.P. (2005). Earthquake energy transmission at interfaces. Proceedings of the 11th international conference of IACMAG, Turin, Italy. Botero E. Mndez B. C., Zapata J. C. and Romo M. P. (2007). Influence of wear on friction variation in concrete sliding interfaces, Proceedings of the 5th International Conference on Seismology and Earthquake Engineering (SEE5), Tehran, Iran, 2007, paper id 147 Cambou B. (1974). Etude du frottement entre matriaux rocheux, Publication E11 of the Institute of Engineering, UNAM, pp. 92 Chaudhuri S. R. and Hutchinson T. C. (2005). Characterizing frictional behavior for use in predicting the seismic response of unattached equipment, Soil Dynamics and Earthquake Engineering, 25(August-October), 591-604. Constantinou M. C., Mokha A., Reinhorn A., (1990). Teflon bearings in base isolation. II: modeling, Journal of Structural Engineering, 116(2), 455-474 Itasca (2005). FLAC3D, Version 3.0. Minneapolis, MI: ITASCA Consulting Group Jangid R. S. (2000). Response of sliding structures to bi-directional excitation, Journal of Sound and Vibration, 243(5), 929 944. Kafali C., Fathali S., Grigoriu M. and Whittaker A. S. (2007). Static and kinetic coefficients of friction for rigid blocks, MCEER Technical Report MCEER-07-0001 Malla S. and Wieland M. (2006). Dynamic stability of detached concrete blocks in arch dam subjected to strong ground shaking, Proceedings of the First European Conference on Earthquake Engineering and Seismology, September 3-8, 2006, Geneva, Switzerland, paper id 1305 Mndez B. C. (2004). Experimental investigation of kinetic friction on a wooden interface. Thesis in partial fulfillment to obtain the Master in Engineering degree, Universidad Nacional Autnoma de Mxico, Mxico D. F. (in Spanish). Mndez B. C. and Romo M. P. (2006). Experiments on frictional behavior of a sliding block. Serie Investigacin y Desarrollo del Instituto de Ingeniera, UNAM, Mxico, DF, 39 pp, ISBN 970-32-3251-5, ISSN 970-32-0196-2 (SID/647) Mndez B. C., Botero E. and Romo M. P. (2006). Frictional approach to seismic analysis of flexible sliding-structures, First European Conference on Earthquake Engineering and Seismology, September 3-8, 2006 Geneva, Switzerland, paper id 682 Naboulsi S. and Nicholas T. (2003). Limitations of the coulomb friction assumption in fretting fatigue analysis, International Journal of Solids and Structures, 40(23), 6497-6512 Newmark N. (1965). Effects of earthquakes on dams and embankments, Geotechnique, 15(2), 139-160. Pekau O. A. and Yuzhu C. (2004). Failure analysis of fractured dams during earthquakes by DEM, Engineering Structures, 26(10), 1483-1502. Rabinowicz E. (1951). The nature of static and kinetic coefficients of friction, Journal of Applied Physics, 22(11), 1373-1379 Yegian M. K. and Lahlaf A. M. (1992). Dynamic interface shear strength properties of geomembranes and geotextiles, Journal of Geotechnical Engineering, 118(5), 760-779 Yegian M. K. and Kadakal A. M. (2004). Foundation isolation for seismic protection using a smooth synthetic liner, Journal of Geotechnical and Geoenvironmental Engineering, 130(11), 1121-1130 Westergaard H.M. (1933). Water pressures on dams during earthquakes, Transactions of the American Society of Civil Engineers, Vol. 98, pp. 418-433