An LSTM-based severity evaluation method for intermittent open faults of an electrical connector under a shock test

The intermittent open fault is the major fault mode in an electrical connector. Therefore, it is necessary to evaluate the severity of intermittent open faults accurately, which would effectively reduce maintenance cost and enhance the useful life. This paper proposes a severity evaluation method based on a Long Short Term Memory (LSTM) network to effectively evaluate the internal looseness degree and the shock acceleration separately. Firstly, a linear model based on the resistance Root Mean Square (RMS) and shock acceleration is constructed. Then the internal looseness feature independent of external shock is extracted. Secondly, the LSTM is used to strengthen the mining of the shape characteristics and evaluate the intermittent open fault severity accurately, whose input is the frequency-domain data of the contact resistance waveform. Results of the comparison verify that the proposed evaluation method not only reduces the number of calculations required, but also improves the severity evaluation accuracy.