断路器机械振动信号特征值提取方法分类性能研究

以ZW45-12型单稳态永磁机构断路器为研究对象,通过搭建数据采集平台得到断路器正常合闸状态、分闸弹簧单根脱落和机构卡涩3种不同运行状态信号数据,利用模糊C聚类分析算法比较短时傅里叶变换、小波包能量熵和希尔伯特—黄变换3种特征提取方法的性能。通过对比研究,得出采用小波包变换方法得到的时频熵向量具有最佳的类可分性能。

Research on Classification Performance of Eigenvalue Extraction Method Based on Circuit Breaker Mechanical Vibration Signal

Circuit breaker mechanical vibration signal is a typical non-stationary signal. This signal contains abundant information of working status and fault feature. The classification performance of its eigenvalue extraction method affects the results of the fault diagnosis directly. In the paper, it takes ZW45 - 12 vacuum circuit breaker with mono-stable permanent magnet mechanism as an example, it has obtained condition data of normal condition of circuit breaker, of opening spring single take off, and of mechanism unsmooth of operating and transmission through construeting the data collection platform. It researches on three timerequency domain signal processing methods of the short-time Fourier transform, wavelet packet energy entropy, Hilbert-Huang transform, adopting the fuzzy C clustering analysis algorithm to compare performance of three kinds of feature extraction methods. Through the comparative study, it concludes that wavelet packet transform method is the best divided performance for the time-frequency entropy vector.