A New Framework for Response Sensitivity Analysis of Soil-Structure Interaction (SSI) Systems

Abstract eng:
Response sensitivities are crucial components for any gradient based optimization process, such as structural reliability, optimization, model updating, and system identification. In the context of finite element (FE) method, this paper presents a new computational framework for response sensitivity analysis based on the direct differentiation method (DDM). The framework can be used for structural, geotechnical or soil-structure interaction (SSI) systems. The DDM-based sensitivity analysis requires differentiating analytically the FE algorithm for the response computation with respect to material\geometry\loading parameters. It provides a semi-analytical solution to the response sensitivities (or gradients) in the sense that the response sensitivities are "exact" once the discretized form (in both space and time domain) of the equation of motion is obtained. The differentiation follows the response computation and involves all hierarchical layers of FE response analysis, namely: (1) structure level, (2) element level, (3) Gauss point level (or section level), and (4) material level. The framework is implemented in a general purpose FE software, OpenSees, an open system for earthquake engineering simulation, which is widely used for static/dynamic, linear/nonlinear FE analysis of structural/geotechnical systems. Base on the newly developed framework, a three dimensional nonlinear frame-soil system subjected to dynamic base excitation is used as benchmark example for studying the sensitivity of material parameters in both structure and soil. In this paper, the sensitivity results provide the relative importance of various material parameters on the responses of the SSI system. To quantify the uncertainty propagation, the sensitivities of both local responses at material/element/section level and global responses at the structural level are studied.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket