Data-driven segmentation of long term condition monitoring data in the presence of heavy-tailed distributed noise with finite-variance

Abstract

Machinery condition prognosis systems use long-term historical data to predict the remaining useful life (RUL). One of the critical steps to reach this purpose is to segment long-term data into two or several degradation stages (healthy, unhealthy, critical stage). Finding a changing point between stages may be a crucial preliminary task for further prediction of degradation process. However, finding the accurate partition into two or more stages is a challenging task in actual application when noise inherent in the observed process exhibits non-Gaussian characteristics. In this paper, a framework for data-driven segmentation is presented for prognosis of machinery long-term data in presence of heavy-tailed distributed noise with finite variance. It is assumed that three different stages are inherent in degradation process and each segment of data follows a specific trend (constant, linear, exponential or polynomial). At first, data is divided into three parts. Trend functions are fitted to the data by using robust regression method, and cumulative error is calculated. This process is done iteratively for all possible partitions into three intervals to find the segmentation which minimizes the error. The framework has been tested via empirical analysis of estimators of the changing points obtained in Monte Carlo simulations. Also, discussed approaches are applied to the real data. In such measurement, data that are commonly available (in condition monitoring systems) is aggregated from the raw signal and sampled at long intervals. Finally, effectiveness of the segmentation results is assessed by comparing them with envelope frequency analysis of raw signal to confirm the fact that detected changing points coincide with start time of the fault in the machine or not.