基于IF和SHAP的无监督机械故障检测和诊断方法

Unsupervised mechanical fault detection and diagnosis method based on IF and SHAP

为实现标签数据不可用情况下对机械故障的早期准确检测和诊断,提出了基于孤立森林(Isolation forest,IF)模型和可解读人工智能工具Shapley加法解释(Shapley additive explanation,SHAP)的无监督旋转机械故障检测方法。提取出时域和频域中的振动特征。在故障检测阶段,以无监督方式基于IF模型检测到早期机械故障。在故障诊断中,使用SHAP对IF模型的预测结果进行解读,通过特征重要性排序得到故障成因,并根据特征与故障位置的对应关系执行故障分类或根因分析。在轴承故障数据集上的试验结果表明,所提方法能够准确及时地检测到早期机械故障的发生,并给出故障发生的原因和位置,所提方法可有效提高机械设备运行和维护的稳定性和自动化。

In order to realize the early detection and accurate diagnosis of mechanical faults without available labeled training data,an unsupervised fault detection and diagnosis method for rotating machinery based on isolation forest(IF)model and explainable artificial intelligence tool Shapley additive explanation(SHAP)is proposed.Firstly,the vibration features in the time and frequency domains are extracted.Secondly,in the fault detection stage,early mechanical faults are detected based on the IF model in an unsupervised manner.Finally,in the fault diagnosis stage,SHAP is used to interpret the prediction results of the IF model.The fault causes are obtained through feature importance ranking,and fault classification or root cause analysis is performed according to the corresponding relationship between features and fault locations.The experimental results on the bearing fault data set show that the proposed method can accurately and timely detect the occurrence of early mechanical faults,the root causes as well as fault locations can be provided accordingly.The proposed method can effectively improve the stability and automation of mechanical equipment operation and maintenance.