Key components identification of EMU complex system faults with interval intuitionistic fuzzy set and multi-attribute group decision-making based on FMECA method

Abstract

With the continuous acceleration of high-speed railway, the high-voltage traction system of the EMU is an important part for ensuring the operation speed and safety. If the failure does not discontinued effectively, it will cause major dangerous accidents, so the key components identification of system is crucial. This paper focus on the contradictions of the expert evaluation information ambiguity, the difference of expert risk appetite and the rationality of risk priority number (RPN) calculation method in the traditional failure analysis method FMECA. The interval intuitionistic fuzzy set (IIFS) is introduced to transform the expert evaluation into the form of membership interval and non-membership interval, which reduced the ambiguity of the specific numerical score. The interval intuitive fuzzy entropy was used to determine the entropy values of the occurrence (O), severity (S), and undetectable degree (D) of each failure mode under each expert score, which was used to calculate the weight value [Formula: see text], to weaken the influence caused by subjective risk preference. The interval intuition fuzzy ensemble operator (AIVIFWM) is used to assemble a single scoring matrix into a comprehensive score, which weakens the subjective influence of expert evaluation. Combined with the multi-attribute group decision-making idea, the score function [Formula: see text] is calculated for each comprehensive evaluation interval of each failure mode after assembly, so as to sort the failure mode risk and finally identify the key components. Based on the fault data of the high-voltage traction system of a certain type of EMU in 2022, 39 failure modes of 30 components are researched and summarized. The results show that rectifier, converter cooling unit, and carbon skateboard are the key components of EMU high-voltage traction system, which provided basic support for the detection and maintenance decision.