An Intermittent Fault Severity Evaluation Method for Electronic Systems Based on LSTM Network

Abstract

Intermittent fault is one of the main causes of the degradation of electronic systems. Accurately evaluating the severity of intermittent faults is a key issue for electronic system fault prediction and health management (PHM). Traditional machine learning methods are difficult to effectively extract the characteristics of intermittent faults. In response to this problem, this paper proposes a method for evaluating the severity of intermittent faults based on LSTM network. This method preprocesses the original data and does not require the process of extracting fault features, then the pre-processed data can be used for the training and testing of the LSTM network. Finally, the paper uses the intermittent fault injector to inject intermittent faults into the key circuits of the electronic system to obtain sufficient fault data to train the LSTM network. The test results show that the proposal are effective and feasible.