Mechanical fault diagnosis method based on migration relation network

Abstract

The invention discloses a mechanical fault diagnosis method based on a migration relationship network. The method comprises the following steps: constructing source domain and target domain data of the migration relationship network; constructing a training set and a test set of migration relationship network samples; constructing a migration relation network capable of detecting the mechanical fault type; and training the migration relation network to obtain a mechanical fault diagnosis model, and performing test and performance evaluation on the final model. The invention provides a migration relationship network with a Siamese structure, which combines a relationship network in meta-learning and migration learning for the first time. A double-channel relation network is constructed by utilizing a Siamese structure, all data of a source domain and label-free data of a target domain are respectively input, information of the target domain is fully considered during additional training, and the accuracy of fault diagnosis is greatly improved. MK-MMD is fused into a network, so that the probability distribution distance between two different fields is effectively reduced, and the application of laboratory data to actual mechanical fault diagnosis becomes possible.

Description

technical field [0001] The invention belongs to the technical field of machine learning, and in particular relates to a mechanical fault diagnosis method based on a migration relational network. Background technique [0002] With the rapid rise and vigorous development of the industrial Internet and the Internet of Things technology, the multi-source sensor network of mechanical equipment is densely arranged, and the amount of monitoring data interaction is increasing day by day, making mechanical fault diagnosis enter the "big data" era, which provides a comprehensive basis for the comprehensive control of equipment. Healthy service status provides big information and big knowledge, and how to effectively tap the potential value behind the big data of machinery has become a frontier hotspot and research difficulty in ensuring the safe operation of equipment driven by big data. [0003] Intelligent fault diagnosis is an important means to ensure the safe operation of mechani...

AI Technical Summary

Problems solved by technology

However, these studies are generally based on the assumption that typical fault information is abundant and health marker information is sufficient

In engineering practice, the above assumptions are difficult to satisfy, because the monitoring data of actual engineering equipment has the following characteristics: ①The value density is low, that is, it is difficult to obtain marked fault data from some machines, and the obtained data has no faults Data occupies the majority

② The availability rate is low. During the long-term operation of the equipment, a large amount of monitoring data has been accumulated, but the data with health marker information that can be used for fault diagnosis is scarce

Current research shows that existing mechanical fault diagnosis knowledge can be used to identify mechanical health status with relevant fault information; reducing the data distribution differences introduced by factors such as variable working conditions and different environmental interferences is the best way to apply transfer learning to intelligent diagnosis of mechanical faults The key; deep learning can adaptively characterize fault characteristics robust to environmental conditions and random disturbances in mechanical monitoring data, which helps to suppress but cannot eliminate data distribution differences

[0006] The current research still has the following problems to be solved: the existing research is limited to the migration diagnosis task of the same machine in different operating conditions or test environments, and the accuracy of deep migration diagnosis between different machines is low. The validity is based on the assumption that the available monitoring data obtained by the diagnosed machinery under a single working condition or test environment is sufficient, which is inconsistent with the monitoring data characteristics of the actual engineering equipment, that is, the lack of fault information and the lack of label information, and it is difficult to adapt to and meet the requirements of mechanical failures. The engineering application requirements of intelligent diagnosis; the existing work has insufficient research on the nature of the knowledge transfer carrier, and ignores the degree of correlation between the source machine and the target machine of the diagnosis knowledge, resulting in unclear fault migration diagnosis mechanism, which restricts the intelligence of mechanical faults Transformation of diagnostic model from individual application to universal application

Images