Constitutive Modeling in Earthquake Simulation Stein Sture and Hon-Yim Ko Department of Civil, Environmental, and Architectural Engineering University of Colorado at Boulder In modern nonlinear coupled motion and pore-water pressure analysis techniques based on Biot's formulation or general theory of mixture concepts, it is important to include realistic yet physically transparent constitutive models, which are capable of accurate simulation of deformation, strength, loading history, nonconservative and dilatancy behavior, as well as mean effective stress dependence during loading, unloading and reloading cycles. While a large number of constitutive models for soils have been developed over the years, many of which are in principle able to simulate behavior, many of these models are very complex, and it is often difficult to relate the models' variables, parameters and constants to basic soil properties. Moreover, validation or verification of model behavior is often lacking. Also, many models are formulated in two-dimensions, such as axisymmetry, and do not readily transform to plane strain or three-dimensions. Perhaps more important, many constitutive models, especially those based on incremental elasto-plasticity are devised in rate form, and little consideration is given to how these models are integrated within nonlinear finite element or difference method implementations of standard virtual displacement equations or Biot's or mixture equations(Jeremic et al., 1997, 1998). These issues, in addition to experimental observations of constitutive behavior, and philosophical discussions are among the topics, which are our presentation centers on. The following paragraphs briefly describe recent and newly resurrected work related to earthquake and constitutive simulations. While nonlinear finite element analysis implementations of Biot's equations focused on formulation of field equations, appropriate kinematic formulations, nonlinear solution algorithms at the system or soil-structure level, and to a lesser extent constitutive model integration, we believe many analysts may be overlooking important physical phenomena, such as localization of deformations in narrow or wide shear bands. While we often observe shear bands in experiments, especially on medium to dense specimens, shear zones also occur in loose specimens. It is now possible to develop finite element analysis technology by means of adaptive or moving meshes, that can incorporate localization phenonena. Shear and related dilatancy behavior clearly result in volume change, and as a consequence undrained shearing can result in substantial excess pore water pressure generation in critical shear zones, which may result liquefaction, flow and failure. It is asserted that this can happen far quicker than predicted via conventional continuum or element-average analyses, in that the shear-zone region is relatively small, and excess pore pressure dissipation into the neighboring soil, which may serve as a sink, happens on a very different time scale than we conventionally consider. While pore water pressures might increase fast in such an analysis, they may also dissipate quickly. Clearly, the development of failure zones in sandy soils, including fracture zones in weakly cemented soils, is strongly linked to interactions between the solid skeleton and water. The interaction, effective stress level, and dilatancy rate will influence the critical load level, where localization takes place as well as the orientation of zones of localized 1 deformation. The initially undrained deformation condition, which governs subsequent behavior, has received considerable attention, and we mention the works of Runesson et al. (1995), Han and Vardoulakis (1995), and Runesson, Larsson and Sture (1998) who studied deformation, strength and instability phenomena for different ranges of pore fluid compressibility. It was found that initial shear band formation, pore-pressure build-up, and soil-structure response are very sensitive to effective stress level, initial density, as well as basic parameters such as friction, dilatancy and elastic moduli. In order to gain further insight to cyclic behavior of sandy soils, Gong (2001) conducted three-dimensional micromechanical computational studies of assemblies of ellipsoidal particles. Uniform or specified size distributions of particle assemblies were pluviated into the analysis domain by means of gravity with specified boundaries or isotropically compressed under either displacement or traction boundary conditions. The investigation included methods for detecting and defining contacts between the ellipsoidal particles. The constitutive behavior at the particle contacts was described by nonlinear elastic Hertz and load-history dependent elastic-plastic slip MindlinDeresiewicz models. A technique for handling both displacement and traction boundaries was developed and is described by Gong (2001). Traction boundary conditions were devised by means of a simulated flexible membrane-system, while the rigid displacement boundary is implemented either in terms of force or displacement control. A new integration algorithm was developed specifically for static analysis of granular materials, which shows increased performance in obtaining accurate and stable positions for all particles in the assembly. The study included numerical simulations of true-triaxial cubical, axisymmetric triaxial-prismatic, direct shear, direct simple shear, pull-out and penetration tests. Numerical experiments were also performed, where the intermediate principal stress (traction) was varied between the major and minor stresses, or where different pure displacement or mixed traction-displacement conditions were applied. The number of ellipsoidal particles present in the various three-dimensional experiments were in the range of 300 to 1,000. The simulations required nearly 30 hours on modern, high-clockspeed and large memory PC-machines. The effects of both smooth and frictional walls (displacement boundaries) were studied. In general, the results compared favorably with data obtained in related laboratory experiments. Sensitivity studies involving varying particle properties (Young's modulus, Poisson's ratio, and Coulomb friction) and time-step sizes are described byGong (2001). It was found that volumetric change and modulus behavior during load cycles are very sensitive to particle shape and thus, fabric. To study behavior of granular soils at low stress levels a set of ground-based and microgravity-based (Space Shuttle) experiments on clean F75 quartz sand were conducted under drained conditions at effective stress levels in the range of 0.05 kPa to 20.0 kPa. Batiste (1998, 2001) and Sture et al.,1998) observed increasing dilatancy angles with decreasing confining stress levels, very high peak friction angles, and very high elastic moduli. Dilatancy angles in the range of 30 degrees were observed for confining stresses in the range of 0.05 to 1.30 kPa for initial relative densities in the range of 65% to 85%, and the dilatancy angles reduced to nearly 20 degrees at 3.5 kPa. Peak friction angles were also extremely high, as one would expect, and these were in the range of 75 to 48 degrees for the same levels and range of initial densities. The data also indicate that the critical void ratio is substantially higher at low stress levels, almost in the range of the 2 maximum void ratio. Triaxial experiments on loose and very loose specimens (10%, 20% and 30%) resulted in substantial volumetric expansion, rather than contraction, even at effective confinement levels in the range of 2.5 kPa. Experiments at confinement levels in the range of 5.0 kPa and higher showed volume contraction during shearing according to standard concepts. Numerical simulations on assemblies of discrete ellipsoidal particles prepared at comparable packings showed similar behavior. The fabric in undeformed and deformed specimens, especially shear localization, were studied using computer tomography (CT) techniques, which showed high accuracy in detecting nonuniformities and shear patterns. Multiple symmetrical radial shear bands were observed in the specimens tested at low stress levels, while nonsymmetrical and far fewer shear bands were observed in specimens subjected to higher stress levels. Void ratio distributions within and outside the shear bands were studied, and it was observed that the variations are of the order of 50%. This again points to shear bands as early generators of excess pore water pressure in the case of initially loose sands, or pore water sinks in the case of dilating bands. Moreover, the shear bands may be serving as internal drainage boundaries. In constitutive modeling of soils a major goal has been to develop cyclic models, which in association with the Biot-field equations, are able to accurately predict build-up of pore water pressure in cycles. In order to achieve this, the models need to include realistic formulations for dilatancy. Nearly 15 years ago the authors developed a "fuzzy set" plasticity theory proposed by Klisinski et al. (1987, 1988), which is based on the mechanical-sublayer concept. Instead of utilizing multiple yield or bounding surfaces a general surface is introduced that span the stress and membership functions used in fuzzy set theory. The model formulation and implemenmtation are relatively simple, in that separate membership functions and related stress spaces are used for the deviatoric and volumetric components. Kinematic plastic behavior is achieved via simple shifting of the apex location of the apex of the pyramidal structure representing the stress and membership function space. Details of this model are contained in Klisinski et al. (1987, 1988). We are continuing the study of this model and its use in nonlinear finite element analysis of long-term cyclic loading, mainly because of its relative simplicity in calculations. Also, we intend to use the experimental observations and micromechanical simulations described above to tune and verify model behavior. An important feature is that the fuzzy set plasticity formulation allows for smooth transitions between the elastic and plastic states, which makes it very advantageous in cyclic simulations. Hysteresis loops and plastic strain increments for successive cycles can be described in a relatively simple manner. The concept allows formulation of several different models within the same mathematical framework, such as models with and without memory, and fading memory, which may be important to simulate variations in soil skeleton behaviro during cyclic loading. Overall verification of fully implemented Biot analysis will be conducted by means of centrifuge model studies. References 3 Alshibli, K.A., Sture, S., Costes, N.C., Frank, M., Lankton, M.R., Batiste, S.N., and Swanson, R.A., (2000), "Assessment of Localized Deformations in sand Using X-Ray Computed Tomography", ASTM, Geotechnical Testing Journal, Vol. 23, No. 3, pp. 274-299. Gong, L. (2001), "Discrete Element Modeling of Three-Dimensional Assemblies of Ellipsoidal Particles", Ph.D. Dissertation, University of Colorado at Boulder, 180. Han, C., and Vardoulakis, I. (1991), "Plane-strain Compression Experiments on Water Saturated Fine-grained sand", Geotechnique, Vol. 41, No. 1, pp. 49-78. Jeremic, B., and Sture, S. (1997), "Implicit Integrations in Elastoplastic Geotechnics", J. Mechanics of Cohesive-Frictional Materials, John Wiley & Sons, 2, 2, 165-183. Jeremic, B., Sture, S., and Runesson, K. (1999), A Model for Pressure Sensitive Materialks Subjected to Large Deformations", Int. J. Solids and Structures, 36, 49014918. Klisinski, M., Alawi, M.M., Sture, S., Ko, H.-Y., and Wood, D.M. (1987), "Elasto-Plastic Model for sand Based on Fuzzy Sets", Proc. Int. Symp. on Constitutive Equations for Granular Non-Cohesive Soils, Ed. A.S. Saada, Publ. A.A. Balkema, Rotterdam, The Netherlands, 1989, 325-348. Klisinski, M., Sture, S., Runesson, K. and Ko, H.-Y. (1988), "Incremental Constitutive Relations for Cohesionless Granular materials Based on Fuzzy Sets", Proc. ASCE Eng. Mech. Conference, Virginia Tech., Blacksburg, 1, 150-152. Runesson, K., Peric, D., and Sture, S. (1996), "Effect of Pore Fluid Compressibility on Localization in Elastic-Plastic Porous Solids under Undrained Conditions", Int. J. Solids and Structures, 33, 10, 1501-1518. Runesson, K., Larsson, R., and Sture, S. (1998), Localization in Hyperelasto-Plastic Porous Solids Subjected to Undrained Conditions, Int. J. Solids and Structures, 35, 4239-4255. Sture, S., Costes, N.C., Batiste, S.N., Lankton, M.R., Alshibli, K.A., Jeremic, B., Swanson, R.A. and Frank, M., (1998), "Mechanics of Granular Materials at Very Low Effective Stresses", ASCE, J. Aerospace Engineering, 11, 3, 67-72. 4