Real-time Data Feature for Health Based on State Change Rate

This paper uses PSLOC (Predictive Sliding Local Outlier Correction) to remove random fluctuations and noise interference. The time series prediction of the system's signals over a period of time is performed based on the least squares support vector machine (LSSVM). The adaptive state change rate is used to detect the prediction starting point. The system's full life cycle signal is divided into several windows, and the state change rate of the signal in each window is calculated. The threshold is determined by calculating the state change rate of the full life cycle signal. Compared with the traditional denoising and smoothing algorithm for non-time series, the algorithm used is more suitable for outlier detection of time series whose trends change over time. The method used in this paper is compared with the traditional wavelet packet threshold denoising method. It is found that the method used in this paper can retain the degradation information of the system to the maximum extent while removing outliers, and is more suitable for data preprocessing for remaining life prediction. The algorithm can not only remove the interference of random fluctuations on state division and remaining life prediction, but also preserve the complete degradation trend of the system adjusted over time.