Reliability design of multibody systems using sample-based extreme value theory

Abstract

A multibody system can be modeled with multiple parameters such as mass, stiffness, damping, and length. Even though such parameters are frequently assumed to be deterministic, they are not because of various factors such as manufacturing tolerance, material irregularity, fatigue, and wear. Because the performance of a multibody system depends on its parameters, parameter uncertainties result in system performance uncertainty. Probability density functions (PDFs) of uncertain parameters can be identified based on their populations. In practical engineering problems, however, it is almost impossible to enumerate the populations of all parameters. Therefore in this study, we propose a sample-based reliability design method using an extreme value theory. The effectiveness and accuracy of the proposed method is validated with three explicit functions and two multibody systems.