Mechanical equipment associated measuring point transfer learning method based on multi-scale convolutional network

Through the multi-scale convolutional network and the maximum mean square difference method, knowledge migration between mechanical equipment related measurement points and target measurement points is achieved, the problems of inconvenient sensor installation and missing data are solved, and the accuracy and generalization of fault diagnosis are improved. sex.

CN119004186AActive Publication Date: 2024-11-22KUNMING UNIV OF SCI & TECH
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411251622.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-08
Publication Date
2024-11-22
Estimated Expiration
2044-09-08

AI Technical Summary

Technical Problem

The complex layout and high load operation of mechanical equipment in industrial sites lead to data loss, limiting the application of deep learning models in fault diagnosis, especially the problems of inconvenient sensor installation and poor knowledge transfer diagnosis capabilities.

Method used

Using the transfer learning method of mechanical equipment related measurement points based on multi-scale convolutional network, the shared network structure is jointly trained through the sliding window mechanism and the maximum mean square difference method, and the characteristics of the related measurement points and target measurement points are extracted to realize the transfer of knowledge. and diagnosis.

Benefits of technology

It effectively improves the generalization and diagnostic accuracy of the model, solves the problems of inconvenient sensor installation and large differences in data distribution, and realizes accurate diagnosis of mechanical equipment failures.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004186A_ABST
    Figure CN119004186A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-scale convolutional network-based mechanical equipment associated measurement point transfer learning method, which comprises the following steps of: firstly, taking source domain data with labels as input of a domain learning network of a source domain to extract associated measurement point features, and sharing a network structure and a weight to the domain learning network of a target domain; taking the target domain data without labels as the input of a domain learning network of a target domain to extract target measuring point features; performing inter-domain difference measurement on the associated measurement point features and the target measurement point features by using a maximum mean square difference method, constructing a joint loss function as an optimization target by combining the classification loss of a domain learning network of a source domain and the inter-domain difference loss, and performing iterative training by using a back propagation algorithm, and target domain learning model parameters are updated, sample categories are predicted, finally, a target domain learning model with an excellent diagnosis effect is obtained, and knowledge migration between associated measurement points and target measurement points of the mechanical equipment is realized. According to the method, deep features in the data are effectively extracted through the multi-scale convolutional neural network, and knowledge migration among different measuring points is realized by using the maximum mean square difference, so that fault diagnosis is performed by using associated measuring points under the conditions that the sensor is inconvenient to install, the data distribution difference is large and label information is lacked, and the fault diagnosis efficiency is improved. And the model generalization is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Mechanical fault diagnosis method based on migration relation network

    CN111709448A

  • Mechanical fault diagnosis method based on multi-sensor information fusion migration network

    CN112161784A