A new method to realize cyclic time averaging for change detection in dynamical machine status

摘要

Detection of changes in dynamical machine status is of great importance to enhance the function of industrial machineries. Cyclic time averaging (CTA) is a powerful tool to achieve this end, but extraction of the so-called average time waveform (ATW) is still challenging. This paper presents a new method to derive the ATW based on optimal warping path which is coupled with adaptive online cycle segmentation. The method has obvious advantages such as no requirement of sensor for calibration and signal pre-processing algorithms in its computation. Taking merit of this method, an automatic analysis of continuous monitoring of vibration signal is proposed by means of a null hypothesis testing to monitor the machine status. The proposed framework is applied to various engineering applications where excellent performance is demonstrated outperforming state of the art methods. Comprehensive experimental results, along with theoretical interpretation and explanation, suggest the great potential of the framework in real applications. (C) 2019 Elsevier Ltd. All rights reserved.