In order to improve the diagnostic accuracy and diagnostic speed of hydropower unit fault diagnosis,this paper proposed an adaptive variational modal decomposition fused with a deep convolutional neural network with a wide convolutional kernel in the first layer for hydropower unit fault diagnosis.Firstly,the sparrow search algorithm was used to optimize the VMD decomposition parameters,and the optimal decomposition parameters were used to decompose the vibration signals of the hydropower unit,so as to achieve the optimal adaptive decomposition of the vibration signals.And then the decomposed IMF components were normalised.Finally,the processed components were inputted into the WDCNN model to be trained and tested,and the results were obtained for the diagnosis of the faults.Comparison experi-ments were carried out with the measured vibration signals of hydropower units.The results show that the proposed method has the optimal diagnostic accuracy as well as good training speed and noise reduction effect,and it has a certain reference role in the actual diagnosis of hydropower unit faults.
维普
万方数据