A Novel Hybrid Switching RUL Prediction Based on SNR Threshold of Linear Motor

This paper aims to solve the problem of predicting the Remaining Useful Life (RUL) of linear motors under winding insulation degradation. First, a new Health Indicator (HI) standard based on the motor model is proposed, and described with Ito process from perspective of stochastic process. Extended Kalman Filter (EKF) is used to observe the degradation of HI. In the early stage of RUL prediction, the basic prediction function is realized through the Instance Based Learning (IBL) method, and Maximum Likelihood Estimate (MLE) method is used in the mid-time and later to improve the accuracy of the prediction. Furthermore, through wavelet denoising of HI, a switching algorithm is available based on Signal-to-Noise Ratio (SNR) as threshold, which overcomes the upper limit of prediction accuracy of IBL. Finally, the effectiveness and reliability of the improved IBL can be verified and tested from simulation.