engleski

RELIABILITY ANALYSIS BASED ON THE STATISTICAL MOMENTS CALCULATION AND FITTING OF THE PERFORMANCE FUNCTION'S PROBABILITY DENSITY FUNCTION

This paper presents a reliability analysis approach which is composed of raw statistical moments calculation and fitting of performance function’s probability density function (PDF). Raw statistical moments are calculated using univariate dimension reduction method (DRM) with Gauss-Hermite integration points. After raw statistical moments are obtained, fitting of probability density function is performed by superposition of normal distributions (SND). Assuming that PDF can be decomposed on a number of different normal distributions, weight of each distribution can be obtained by solving linear system equations that is made of the raw statistical moments of each normal distribution. Nelder-Mead search algorithm is used for final refinement of the PDF. A few numerical examples are presented and obtained results are compared with the Monte Carlo simulation (MCS) and first order reliability method (FORM). The method, due to small number of system evaluations needed, can be very useful for the multicriterial design processes using reliability/risk based design attributes.

reliability analysis; dimension reduction method; superposition of normal distributions

nije evidentirano