Speed reducer fault analysis method based on neural network

The invention relates to the technical field of fault detection, in particular to a speed reducer fault analysis method based on a neural network, and the method comprises the steps: obtaining a plurality of types of fault data sequences of a to-be-detected speed reducer; obtaining a feature scale of each kind of monitoring sample data of the target fault; acquiring a fault response coefficient of each kind of monitoring sample data; sorting all kinds of monitoring sample data according to the fault response coefficient to obtain a transfer sequence of the target fault; according to the transmission loss degree and the feature scale of each type of monitoring sample data, obtaining a fusion weight of each type of monitoring sample data of the target fault; training a CNN auto-encoder according to the fusion weight of the monitoring sample data; and performing fault prediction on the speed reducer to be detected according to the trained CNN auto-encoder. The accuracy of the fault detection result of the speed