Recurrent Neural Network-Based Fault Diagnosis Method for Multisource Information Fusion of Drilling Pump Power End Gear

Drilling pump is an essential mechanical equipment for oil and gas exploitation. The existing gear fault diagnosis technology is difficult to effectively diagnose the fault state of drilling pumps, which are complex nonlinear high-dimensional coupling systems. Most data-driven methods ignore the time sequence characteristics in the signal. This paper introduces a recurrent neural network (RNN) to process the time sequence data extracted from gear fault vibration processes. It adopts stacked denoising autoencoders (SDAE) and proposes a new data sampling method to improve the generalization ability and anti-noise interference ability of the model proposed in this paper. Moreover, 17 groups of signals were collected for four working conditions and five health states through the test bench. The collected data were used to verify the model proposed in this paper. The results show that the proposed method can effectively capture the fault characteristics of gears under different working conditions, providing an effective solution for intelligent fault diagnosis methods of drilling pump gears.