The performance degradation process of the airborne fuel pump has of multi-stage and nonlinear characteristics, which requires real-time life prediction. To address these issues, an online degradation model and a life prediction method based on failure physics and data driven are proposed. The fuel pump degradation stage is identified online by the switching Kalman filter, the degradation model of rapid degradation stage is formulated based on failure physics and data-driven method, the model parameters are continuously updated based on the unscented Kalman filter, and the failure life is predicted by using the updated model. The proposed method is compared with the data-driven method, the fusion method without degradation stage identification or parameters update. The root mean square value is less than 0. 3 during the whole parameter update process, and the percentage error of lifetime prediction is less than 2% , which are smaller than the values of the compared method. The effectiveness and superiority of the proposed method are verified.