Seismic site response analysis (SSRA) is typically performed considering only one horizontal component of earthquake excitation. In many cases, however, two or three components are needed for the analysis to properly account for the true multidirectional nature of seismic loading. In this type of analysis, it is essential to use multiaxial constitutive models that can realistically describe the stress-strain response of soils. Development and validation of such constitutive models are essential steps toward this goal. Fortunately, a large quantity of experimental data from multidirectional cyclic shear tests is available and can provide physical basis for validating such models. This paper focuses on evaluation of two members of the SANICLAY and SANISAND families of constitutive models for simulating the response of clay and sand, respectively, when subjected to multidirectional cyclic shearing. The models have anisotropic elasto-plastic formulation, within the framework of critical state soil mechanics, and follow the bounding surface plasticity theory. They are calibrated and evaluated against experimental data on Gulf of Mexico clay and Monterey No. 0/30 sand in undrained multidirectional cyclic shear tests, including linear, circular/oval, and figure-8 loading paths. This study provides a basis for evaluation of the capabilities of these models in multidirectional shearing, thereby paving the way towards future applications in multidirectional SSRA.

Introduction

In recent decades, the destructive nature of earthquakes on constructed facilities has led to extensive research focused on the dynamic properties of soil. Direct simple shear (DSS) test has been chosen as a close configuration to model the plane strain condition and the rotation of principal stress axes in soil [1]. In this test, during the cyclic regime only one horizontal shear component is exerted on the specimen. With or without an offset static shear stress in the same direction as the cyclic shear stress, this unidirectional shear mode can be used to replicate the response of soil subjected to one-dimensional propagation of shear waves. In the field, however, shear wave propagation is multidirectional. Even if the vertical component of the seismic loading is neglected, there exist two horizontal shear components as depicted in Fig. 1(a), and neglecting one of them can potentially lead to underestimation of seismic demand. To mimic the response of soil element under level or sloping grounds, when subjected to multidirectional cyclic shearing, a number of more sophisticated devices for simulating the multidirectional cyclic shearing have been established, developed and refined over the years [2,1,3–۷]. These apparatuses are very useful in generating a comprehensive experimental database for evaluation of various constitutive models in such complex loadings. The soil sample in a multidirectional cyclic shear test goes through two loading stages. The first one, referred to as consolidation stage, is to reproduce the corresponding in situ state of soil cconsolidated under level or sloping grounds as shown in Fig. 1(a). For modeling the initial condition of soil under a level ground (away from a slope), a soil element is consolidated vertically with the lateral normal strains constrained; this is typically referred to as K0 condition.