Self-improving Method for Health State Identification Network of an Electromechanical Composite Transmission System

Abstract

The present invention belongs to the field of intelligent operation and maintenance of mechanical equipment, and discloses a self - improvement method for a health state identification network of an electromechanical composite drive system, including the following steps: S1: Training the health state identification network using a data set to obtain weight parameters; S2: Freezing the weight matrix of the feature extraction layer and obtaining distal memory samples based on the data set; S3: Using the distal memory samples to perform partition conversion on the weight matrix of the linear layer to obtain a reconstructed activation region and a reconstructed dormant region; S4: Freezing the reconstructed activation region and updating the reconstructed dormant region using the data set to obtain the weight matrix of the linear layer; Obtaining distal memory samples based on the data set; S5: Continuously repeating S3 - S4 to obtain a self - improved health state identification network. The present invention can improve the problem of catastrophic forgetting, avoid the problem that few - sample data is prone to overfitting, and enable the health state identification network to have good identification accuracy.

Images