基于PCA与决策树的转子故障诊断

ROTOR FAULT DIAGNOSIS BASED ON PCA AND DECISION TREE

将数据挖掘方法引入旋转机械故障诊断领域,提出一种基于主元分析(PCA)与决策树相结合的转子故障诊断方法。该方法首先利用PCA进行特征约简,降低特征空间的维数,然后采用C4.5决策树进行训练学习以及诊断决策。通过对转子类常见故障的诊断分析,证明该方法具有比BP神经网络训练时间更短、诊断准确率稍高的特点。

A method of rotor fault diagnosis based on principal component analysis (PCA)and decision tree is proposed. With this method, PCA is used to simplify rotor fault features and decrease the dimension numbr of the feature space. Then the decision tree C4.5 is applied to learn from training samples and diagnose by using the acquired knowledge. Through the diagnosis test of five rotor faults, it is proved that C4.5 decision tree needs less training time and has higher correctness rate than back propagation (BP) neural networks.