Remaining Useful Life Estimation of Aero-Engines with Self-Joint Prediction of Continuous and Discrete States
share
The remaining useful life (RUL) estimation generally suffer from this problem of lacking prior knowledge to predefine the exact failure thresholds for machinery operating in dynamic environments. In this case, dynamic thresholds depicted by discrete states are effective to estimate the RUL of dynamic machinery. Currently, only few work considers the dynamic thresholds, and adopt different algorithms to predict the continuous and discrete states separately, which largely increases the complexity of the learning process. In this paper, we propose a novel prognostics approach for RUL estimation of aero-engines with self-joint prediction of continuous and discrete states within one learning framework. With modeling capability of self-adapting structure and parameters online, the quantized kernel recursive least squares (QKRLS) algorithm is introduced to predict the degrading states and also determine the discrete states with the kernel centers. The self-evolving dynamic kernel centers obtained during building predictors are automatically assigned as the discrete states for different engines without predefining them. Then, the RUL is estimated conveniently once the predicted degrading signals fall into the final fault state based on a distance metric. Finally, the results from turbofan engine datasets demonstrate the superiority of the proposed approach compared to other popular approaches.
READ FULL TEXT
