다봉분포를 고려한 신뢰성 기반 최적설계

Abstract

In recent studies of reliability analysis methods, estimation of reliability for unimodal and bimodal distribution has been done separately. However, in reliability-based design optimization, the number of output modes depends on input modes, design point, function nonlinearity, and so on. Therefore, in this study, we define a probability density function (PDF) in exponential polynomial form for multimodal distribution. Parameters for the multimodal PDF can be estimated by maximum likelihood estimation and then effective polynomial terms can be determined by Akaike information criterion. Finally, distribution of each design point is estimated as exponential polynomial form and the reliability is calculated by integration of the PDF.