Harmonic reducer fault diagnosis method and system based on generative adversarial network

Abstract

The invention relates to a harmonic reducer fault diagnosis method and system based on a generative adversarial network. The method comprises the following steps: S1, carrying out data preprocessing, collecting vibration acceleration signals of a harmonic reducer, extracting original signal features, and constructing an original data set by using normalized data; S2, carrying out data generation, and generating various types of fault data by utilizing a plurality of generative adversarial networks; S3, performing data selection, filtering and purifying the generated data by using a data selection module, and performing screening; and S4, carrying out fault classification to form a new balanced data set, and using the multi-scale convolutional neural network as a classifier to carry out multi-classification of harmonic reducer faults. The high-quality fault data of the harmonic reducer is generated through the generative adversarial network, the balanced data set is constructed together with the real data, and the multi-scale convolutional neural network is used for fault diagnosis, so that the multi-classification precision of the harmonic reducer is improved under the condition of data imbalance.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis and online monitoring of industrial robots, in particular to a fault diagnosis method and system for a harmonic reducer based on a generative confrontation network. Background technique [0002] Harmonic reducers are widely used in industrial robots due to the advantages of high transmission ratio, no backlash, high compactness and lightness, good resolution and excellent repeatability, to increase transmission to or from a shaft end The amount of torque transmitted by the shaft end. The complex design of the harmonic reducer is sensitive to manufacturing and assembly errors, and abnormal vibrations are related to operating conditions. Even small errors can cause excessive vibrations that can damage the performance of the robot. In addition, a harmonic reducer is a highly nonlinear system that is usually coupled with other external electromechanical systems, so vibration signals often exh...

AI Technical Summary

Problems solved by technology

The existing literature only studies the components of the harmonic reducer, such as reduction gears and bearings, and is not aimed at the finished product of the harmonic reducer as a whole, so it is difficult to reflect the real operation of the entire harmonic reducer, and the current Some time-frequency domain analysis methods, the diagnostic effect needs to be improved

Various machine learning methods based on data balance can obtain high classification accuracy in the fault diagnosis of mechanical equipment such as roller bearings, but they must have a large amount of balanced data of various types, so this is not suitable for harmonics reducer

Images