Long-term gear life prediction based on ordered neurons LSTM neural networks

摘要

Gear failure may affect the operation of mechanical equipment, and even cause the catastrophic break of machine and even casualties. Thus, the remaining useful life (RUL) estimation of the gear has important significance. This paper proposes a gear RUL prediction model based on ordered neurons long short-term memory (ON-LSTM) networks. The proposed methodology consists of two parts: firstly, extract the health index by computing frequency-domain features of raw signals; secondly, the ON-LSTM network model is constructed for generating the target output of the RUL prediction. Unlike the traditional LSTM neural network, the developed model integrating tree structures into LSTM to use the sequential information between neurons, so it has enhanced predictive ability. In comparative experiments, the Scores of ON-LSTM, LSTM, GRU, DLSTM and DNN are 0.398, 0.129, 0.07, 0.029 and 0.102, respectively; and ON-LSTM successfully fulfils twenty-three tasks while LSTM just fulfils five tasks in long-term prediction. Moreover, ON-LSTN only requires about 400 iterations for convergence, which is much faster than other RNNs. Experimental results show that ON-LSTM network achieved the best accuracy of short-term and long-term prediction, and it has the best robustness and convergence speed. And it can be effectively applied to the RUL prediction of gears. (C) 2020 Elsevier Ltd. All rights reserved.