Cross-device diagnosis method based on multi-source domain data knowledge, electronic device and storage medium

Through a cross-device diagnostic method based on multi-source domain data knowledge, using iso-window division and collaborative training technology, combined with the adversarial structure of the custom threshold equalization module, the data distribution differences and sample scarcity problems in cross-device diagnosis are solved, and the diagnosis is improved Accuracy and robustness.

CN119249259APending Publication Date: 2025-01-03SHANGHAI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411090585.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2025-01-03

AI Technical Summary

Technical Problem

In cross-device diagnostic tasks, traditional neural networks are prone to negative migration when passing diagnostic knowledge between different devices, and the lack of sufficient training samples will lead to network overfitting, making it difficult to effectively deal with data distribution differences and operating conditions between different devices Changes lead to poor diagnosis.

Method used

Using a cross-device diagnostic method based on multi-source domain data knowledge, the vibration signal is divided by iso-window, a network model of multi-layer residual block structure and Softmax classifier is constructed, collaborative training is carried out to reduce domain differences, and a custom threshold equalization module is used. Build adversarial structures and optimize network parameters to improve robustness and feature extraction capabilities.

Benefits of technology

Effectively alleviates the impact of the network on the differences in sample characteristics in different fields, improves the robustness and accuracy of the model in cross-device diagnostic tasks, reduces the negative migration error rate, and realizes accurate analysis and identification of bearing acceleration and sound signals. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119249259A_ABST
    Figure CN119249259A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of cross-equipment fault diagnosis, and particularly discloses a cross-equipment diagnosis method based on multi-source domain data knowledge, electronic equipment and a storage medium, and the method comprises the following steps: S1, collecting vibration signals of a bearing in various states under different working conditions; s2, dividing the vibration signal into a certain number of samples by equal window division; s3, constructing a network model based on cooperative training of multiple groups of source domain data, wherein the network model is used for cross-device diagnosis of samples; the network model comprises a multi-layer residual block structure with the same network structure and three independent task classifiers; the residual block structure constructs an adversarial structure through a self-defined threshold balancing module and is used for improving network characterization performance; the classifier is used for identifying fuzzy fault samples in the same category of data of multiple devices.
Need to check novelty before this filing date? Find Prior Art