Research on Prediction Method of Gear Pump Remaining Useful Life Based on DCAE and Bi-LSTM

As a hydraulic pump is the power source of system, predicting its remaining useful life (RUL) can effectively improve operating efficiency system and reduce incidence failure. This paper presents scheme for RUL (gear pump) through combination deep convolutional autoencoder (DCAE) bidirectional long short-term memory (Bi-LSTM) network. The vibration data were characterized by DCAE, health indicator (HI) was constructed modeled to determine degradation state gear pump. DCAE typical symmetric neural network, which extract characteristics from using symmetry encoding network decoding After processing original segment, indicators entered as label into prediction model based on Bi-LSTM training carried out achieve To verify validity methodology, accelerated experiment out, whole cycle obtained method validation. results show that HI characterize degenerative pump, proposed predict degeneration trend