RELIABILITY BASED DESIGN OF GEOTECHNICAL SYSTEMS BASED ON A DECOUPLED RELIABILITY ANALYSIS AND OPTIMIZATION APPROACH

Abstract

Reliability-based design intends to design an engineering system subject to probabilistic design constraints. This paper presents a reliability design optimization approach of geotechnical engineering problems based on a decoupled reliability analysis and optimization concept. Geotechnical and foundation systems involve considerable uncertainties in soil properties, loading conditions, and inadequate site characterization. A traditional reliability design optimization method, such as the reliability index approach, requires a double-loop optimization structure, since reliability analysis is implemented for constraint function evaluation using a numerical optimization algorithm. The decoupled approach separates the reliability analyses of the constraint functions from the design optimization loop. An adaptive metamodeling method is proposed to create approximate functions of the reliability indices, so that the reliability design optimization is performed using the approximate functions of the reliability constraints. Radial basis functions are implemented to build the metamodels. To progressively improve the accuracy of predicting the reliability indices, an adaptive sampling technique is employed, so that new samples are added using the optimal points found in previous design iterations. Numerical examples are solved and presented, and the proposed design approach works well. It is shown to be a useful alternative for solving reliability-based design of geotechnical and foundation systems.