早期故障检测的特征集空间的距离判别法

提出早期故障检测的特征集空间的距离判别法,以含有早期微弱缺陷的元件为参考样本,作为特征集空间的原点,用多特征总体评价的诊断方法,建立特征集空间距离和故障大小的一种对应,最后根据经验预设阈值对检测结果做出定性判断.多特征总体评价,从不同角度充分挖掘早期故障信号中蕴藏的信息,避免片面和盲目.对各个特征量的加权反映出各个特征量在总体评价中重要性差别.故障特征提取用到了分形特征及小波多分辨分析.

Comprehensive Evaluation Approach to Fault Detection Based on Characteristic Set Space

A comprehensive approach to fault detection was proposed. This approach uses weighted characteristic departure values that are extracted from process signals to judge occurrence of fault in a given system. The relevancy between the distance of characteristic set space and degrees of fault is built. Comprehensive consideration of multi-characteristics can avoid unilateral conclusion. The difference of weights distributed to different characteristic values reflects different significance in comprehensive consideration. Qualitative conclusion can be drawn when compared to a given threshold value. Wavelet analysis can provide local information that cannot be obtained using Fourier analysis and statistical estimation theory. To extract fault features from testing signals,Wavelet Multi-Resolution Analysis and fractal analysis were employed.