An intelligent FMEA system implemented with a hierarchy of back-propagation neural networks

Abstract

This paper has used a series of back-progation neural networks (BPNs) to form a hierarchical framework adequate for the implementation of an intelligent FMEA (failure modes and effects analysis) system. Its aim is to apply this novel system as a tool to assist the reliability design required for preventing failures occurred in the operating periods of a system The hierarchical structure upgrades the classical statistic off-line FMEA performance. From the simulated experiments of the proposed BPN-based FMEA system (N-FMEA), it has found that the accuracy of the failure modes classification and the reliability calculation are knowledgeable and potential for performing pragmatic preventive maintenance activities. As a result, this paper conducts an effective FMEA process and contributes to help FMEA working teams to reduce their working loading, shorten design time and ensure system operating success.