Worst-Case Tolerance Analysis of Non-Linear Systems Using Evolutionary Algorithms

In this paper an evolutionary algorithm-based approach to the worst-case analysis of non-linear systems is presented. The evolutionary algorithm is used to minimize the underestimation error which affects classical Monte Carlo-based methods. The method seems to be useful especially whenever large parameter variations need to be taken into account and performance functions that are nonlinear with respect to parameters are considered. An application example involving a lumped non-linear circuit modeling a cable termination equipped with a stress control tube is used to introduce the method.