Predicting Reliability of Structural Systems Using Classification Method

Abstract

This research examines classification approaches for estimating the reliability of structural systems. To validate the accuracy and efficiency of the classification methods, a practical engineering problem; namely, a spider assembly of a washing machine, has been considered. For the spider assembly, fatigue life test, finite element analysis, physical experimentation, and a classification processes are conducted in order to establish the analytical certification of its current design. Specifically, the finite element analysis and fatigue life analysis are performed and their results are validated compared to physical experimental results. The classification process is developed to estimate the probability of failure of the spider assembly in terms of stress and fatigue life. The relationship between the random quantities and structural responses of the spider assembly is established using probabilistic neural network and the support vector machine classifiers. The performance margin of the spider assembly is fully identified based on the estimated failure probability and structural analysis results from the fatigue life analysis and classifications.