Using Parallel Computer Systems to Examine Seismic Reliability of Structures

Gu J

doi:10.4172/2472-0437.1000101

Abstract

Structural response under seismic loadings is typically nonlinear and related to many factors, such as structural configurations, material properties, occupancy loads, earthquake hazards and incomplete knowledge of the system. As all these factors have their sources of uncertainties, structural response under seismic loading has its probabilistic nature. Therefore, the random variable for any structural demand follows a multivariate probability distribution over the integration domain defined by the limit states. Examining the probabilistic behaviour of structures under earthquake loadings has to consider the sources of uncertainties from all factors. It is also known that numerical methods, such as the finite element method, are commonly used to predict nonlinear structural response. The probabilistic structural demand is a discrete probability function of its related variables. In order to examine seismic risks and mitigate potential damages to structures, it is important to accurately quantify seismic reliability of structures. The traditional seismic reliability analysis uses approximate algebra equations with parameters obtained from aggregation of data points of dynamic analysis, which may not be able to produce accurate results. In this paper, probabilistic seismic demands are solved with numerical procedures of the traditional SAC method and the Monte Carlo simulation. These methods rely on the results from repeatable nonlinear dynamic analyses, which were traditionally considered to be a bottle-neck due to limited computing resources. The recent progress in parallel computing technology and open-source software has made such scientific computation affordable for the engineering community. Two parallel computer systems were used to analyze seismic reliability of the structures. One system is based on multiple personal computers in typical computer labs. The other system is to use high performance computer clusters. Both systems were applied to analyze a two–storey wood frame building and a three-storey steel moment building, respectively

Highlights

Structural dynamic response under seismic loading are nonlinear functions of many factors, such as structural configurations, material properties, occupancy loads, earthquake hazards and incomplete knowledge of the system
The random variable for any structural demand follows a multivariate probability distribution for all related factors over the integration domain defined by the limit states
The steel moment frame building was modeled as a “M2” model (FEMA 355C) with zero-length rotational springs to represent plastic hinges and elements with rigid boundaries to represent panel zones

Summary

Introduction

Structural dynamic response under seismic loading are nonlinear functions of many factors, such as structural configurations, material properties, occupancy loads, earthquake hazards and incomplete knowledge of the system. The random variable for any structural demand follows a multivariate probability distribution for all related factors over the integration domain defined by the limit states. The fragility analysis determines the exceeding probability of demand conditioned on a specific level of intensity measure [1,2,3,4,5,6]. The occurrence probability of earthquake intensity measure (IM) is determined by seismologists on a regional basis. Determined hazard levels, such as those specified in the building codes (i.e., the design intensity at 2% in 50 years) are commonly used by engineers. With the determined intensity targets, the fragility analysis provides reasonable information about the probabilistic behaviour of structures

Methods

Results

Conclusion