Hybrid model-driven and data-driven approach for the health assessment of axial piston pumps

Axial piston pumps are key components in hydraulic systems and their performance significantly affects the efficiency and reliability of hydraulic systems. Many data-driven approaches have been applied to the fault diagnosis of axial piston pumps. However, few studies focus on the performance degradation assessment that plays an important role in the predictive maintenance for axial piston pumps. This paper proposes a hybrid model-driven and data-driven approach to assess the health status of axial piston pumps. A physical flow loss model is established to solve for the flow loss coefficients of the axial piston pump under different operating conditions. The flow loss coefficients act as feature vectors to train a support vector data description (SVDD) model. A health indicator based on SVDD is put forward to quantitatively assess the pump health status. Experimental results under different pump health conditions confirm the effectiveness of the proposed method.