Nonparametric reliability-based design optimization using sign test on limited discrete information

Abstract

Many methods for reliability analysis and reliability-based design optimization have been developed since the last decades. However, for most of these methods predicting the reliability, stochastic information of the variables has to be assumed as parameters like the mean (location parameter) and variance (scale parameter). This assumption cannot guarantee the accuracy of the reliability when information is limited. In this paper, we propose a nonparametric RBDO using sign test that does not consider the parameter but requires limited discrete information, like only sample data. We define the uncertainty of the reliability as the decision error of nonparametric hypothesis test due to limited information. We examine the tendency of the solution with respect to the reliability of a system and the uncertainty of the reliability through an example.