<p>Axial piston pump is the primary energy component of a hydraulic system, and evaluating its status is essential for guaranteeing secure functioning. The performance deterioration process in axial piston pump exhibits multistage and nonlinear characteristics. To solve these problems, this study introduces a method for online degradation modelling and the life prediction of axial piston pump based on a data-driven and failure physics fusion. Firstly, a feature fusion strategy is proposed to fuse different types of information and create an health index with trend prediction. Secondly, a degradation model for the rapid degradation stage was constructed based on data-driven and failure physics methods. Finally, the experimental data were used to continuously update the parameters of the built degradation model through an unscented Kalman filter, and the defect lifetime was predicted based on the model after updating the parameters. The proposed approach was compared with both a purely data-driven approach and a fusion method that lacks degradation stage identification and relies solely on parameter updates. The results indicate that life prediction errors were smaller than those of the comparative methods and that the comprehensive scores exceeded those of the comparative methods, thus demonstrating the effectiveness of the proposed method.</p>