A Dynamic Current Feature Map for Condition Monitoring of Rotating Machinery

This paper is devoted to proposing an efficient and visual method for engineers to evaluate the current signals of rotating machinery. Feature extraction methods based on time and frequency domain are applied to define the features of one-phase current signal. In order to dynamically estimate the stability and discover the potential hazards within three-phase current signals, Amplitude Variation Rate (AVR), Phase Fluctuation Rate (PFR) and Shape Inconsistent Rate (SIR) are proposed in this paper as conjoint phase features. A well-organized and strongly related visualization way to indicate the status and deviation degree of rotating machinery called Dynamic Current Features Map (DCFM) is established, where an unsupervised machine learning method is adopted to describe the healthy area. The new map expresses character including rotating speed, time track and the deviation degree from normal condition simultaneously, which can be an appropriate fashion for quick and accurate decision making. Historical data of current signals from healthy motorized spindle, current signals from aged sample and fault signals of motorized spindle are collected to test the availability and efficiency of DCFM and the result proves the map's potential to fulfill the condition monitoring task.