基于不平衡样本特征的电缆发热故障自动化诊断技术OA

In the automatic diagnosis of cable heating fault,the normal operation of cable data is much more than the fault data,resulting in the traditional fault diagnosis model is prone to bias the majority of classes(normal classes)while ignoring the minority of classes(fault classes)during training,which reduces the accuracy of fault identification.Therefore,an automatic fault diagnosis technique for cable heating based on unbalanced sample characteristics is proposed.By extracting the characteristic vector of cable heating fault,XGBoost decision tree is used to predict the unbalanced sample collection required for cable heating fault diagnosis.The BAgging-heterogeneous K-nearest neighbor integration algorithm is used to plan and classify the trained feature set,and a feature set classifier is established,and the automatic diagnosis of cable heating fault is realized according to the classification accuracy weight matrix.The experimental results show that the loss function of the proposed method rapidly converges and remains stable during the training process,which can accurately capture and judge the heating fault state of the cable.The F1 value obtained by the calculation is above 0.95,which further verifies the effectiveness and high precision of the method in the field of cable fault diagnosis.