Detection Method of Transformer Mechanical Fault Based on Mode Decomposition of Vibration Signal
The traditional transformer mechanical fault detection method can only monitor the temperature state after the fault occurs,and the training error is large,so a transformer mechanical fault detection method based on the mode decomposition of vibration signal mode is designed.By calculating the local gradient val-ue,we understand how much each feature contributes to the fault classification.Improving the loss function can improve the performance and accuracy of the model,and establishing the transformer hotspot temperature model can monitor the temperature status in real time and prevent potential faults.Based on the decomposi-tion of vibration signal mode,the parameter characteristics of fault nodes are extracted to identify and diag-nose faults more accurately.Finally,by positioning the location of mechanical failure of transformer,mainte-nance measures can be taken quickly,which provides important technical support for the fault diagnosis of transformer.The experimental results show that the training error of the designed transformer mechanical fault detection method based on vibration signal mode decomposition is kept at 0.05,which proves that the design method is good and has certain research value.
万方数据
维普
