基于特征相关性和冗余性分析的机械故障特征选择研究

Research on Mechanical Fault Feature Selection Based on Feature Relevancy and Redundancy Analysis

从特征相关性和冗余性的定义出发,利用特征与类别间的互信息对特征相关性和冗余性进行了度量,提出了一种基于特征相关性和冗余性分析的特征选择方法。数值仿真和柴油机故障特征选择实验结果表明,新方法可以快速、有效地求得优化特征集,是求解特征选择问题的一个较好方案。

Many features in original fault feature set are irrelevant or redundant to the fault states in mechanical fault diagnosis. These features would decrease diagnosis precision and efficiency. Mutu al information was applied to measure the feature relevancy and redundancy, and a new feature selec tion method based on relevancy and redundancy analysis was proposed. According to the results of data simulation and Diesel engine fault feature selection example, it is proved that this scheme can get optimal feature subset effectively and quickly. The method has good prospects in the fault feature selection.