基于G-DPSO算法的决策树轴承故障诊断方法

Fault Diagnosis of Bearing Based on G-DPSO and Decision Tree

针对机械设备状态监测与故障诊断技术中特征提取对诊断准确性的局限,从原始故障信号数据中提取出尽可能多的有用信息。提出通过最佳特征数据集对轴承故障进行诊断分析,分别从幅域和频域对故障数据进行特征提取。采用一种改善的粒子群(G-DPSO)算法对提取的特征数据集进行筛选,对传统粒子群算法权重系数进行优化,同时和故障诊断需要的决策树模型的信息熵增相结合,可以达到将最适合故障诊断的特征向量提取出来的目的。用5种轴承故障数据对所提方法进行实验分析,诊断正确率能达到97%之上,证明所提出的方法是有效、可靠的。

In view of the limitation of feature extraction in condition monitoring and fault diagnosis technology of mechanical equipment on the accuracy of diagnosis,as much useful information as possible can be extracted from the original fault signal data.It is proposed to diagnose and analyze the bearing fault by using the best feature data set,and extract the feature from the fault data in the amplitude and frequency domains,respectively.An improved particle swarm optimization(G-DPSO)algorithm is used to screen the extracted feature data sets,optimize the weight coefficients of the traditional particle swarm optimization algorithm,and combine it with the information entropy increase of decision tree model for fault diagnosis.It can extract the most suitable feature vectors for fault diagnosis.Five kinds of bearing fault data are used to test and analyze the proposed method.The diagnostic accuracy can reach above 97%,which proves that the proposed method is effective and reliable.