Physics-Based Machine Learning Hydraulic Piston Pump Performance Degradation and Prognostics

The hydraulic piston pump is an important component in the hydraulic system, and it has always been an important problem to accurately and effectively estimate the degradation state of the pump. Physical model-based and data-driven fusion methods can overcome the shortcomings of the two models, and have achieved success in the fields of prognosis and health management. This paper proposes a physics-based deep learning framework for degradation assessment and remaining useful life (RUL) prediction of hydraulic piston pumps. Firstly, the failure physics of the three friction pairs of the hydraulic piston pump is analyzed, and the mathematical calculation formula of the wear amount is obtained. Then, the calculated wear of the three friction pairs is used as the input with physical information, and is fused with the long short term memory (LSTM) network. Finally, the effectiveness of the proposed method is verified by the data of the hydraulic piston pump. It shows that the proposed framework has good capability of degradation assessment and RUL prediction.