关键词: 故障诊断, 数据驱动, 数据预处理, Gap度量, 主元分析, 变化率变换, 特征提取

Abstract:

The method based on deep learning has made great progress and good results in solving small faults, but the prerequisite for sufficient sample data is difficult to achieve in the current situation. So there is still a good need for the fault diagnosis method based on traditional data preprocessing. Principal component analysis(PCA) is widely used in fault diagnosis. Because traditional data preprocessing methods use the absolute distance between samples as the criterion for fault detection and fault diagnosis, the feature extraction is not accurate. For this reason, this paper proposes a data preprocessing method based on rate-of-change(ROC) transformation to improve the performance index of PCA in fault diagnosis. After the original data set is preprocessed by the change rate transformation, it can effectively detect the minor faults in the system variables. Finally, the feasibility and effectiveness of the PCA fault diagnosis method based on the rate of data change are verified by simulation.

Key words: fault diagnosis, data driven, data preprocessing, gap metric, principal component analysis, rate of change transformation, feature extraction