Design and analysis of a new feature optimisation method for circuit breakers defect identification

Abstract

The vibration signals of a circuit breaker (CB) contain important action timing information. The optimisation of features extraction for vibration signals generated during the operation process of CBs is crucial for rapid defect location and identification for CB. The authors propose a new feature optimisation method, defined as energy trajectory entropy of vibration signals via tracing the action characteristics of the main shaft of CBs. Firstly, an image tracking algorithm is employed to dynamically capture the key frames of the high-speed image sequence of the main shaft and accurately divide the action sequence. The “cluster” instantaneous energy waveforms of the vibration signal, divided by zones, are then diffused by the twiddle factor in the polar coordinate sub grid area, while the energy trajectory entropy algorithm (ETE) is utilised to investigate the subtle changes in the energy release process. The ETE scale parameters are optimised using a support vector machine identification model. This enables rapid location of defective components in CBs, resulting in a significant reduction in time cost. The experiment has confirmed the strong feature correlation between CBs action and vibration signals, offering new ideas for achieving non-invasive defect identification of CBs.