Mechanical fault diagnosis method and system based on federal domain generalization

A technology for mechanical faults and diagnosis methods, applied in neural learning methods, machine learning, computer parts and other directions, can solve problems such as personalized training of models that do not consider the availability of target domain data, and achieve cross-domain fault diagnosis problems and reduce Differences, data security and the effect of improving the correct rate of fault diagnosis

Active Publication Date: 2022-07-29
SHANDONG UNIV
View PDF7 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the existing methods assume that the target domain data exists and participate in the training process, without considering the unavailability of the target domain data and the personalized training of the model

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Mechanical fault diagnosis method and system based on federal domain generalization
  • Mechanical fault diagnosis method and system based on federal domain generalization
  • Mechanical fault diagnosis method and system based on federal domain generalization

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0048] Let N clients contain N source domain datasets , the N+1th client has the target domain dataset ,in, represents the total number of samples in the target domain dataset, means the first A sample of the target domain dataset, The value range is from 1 to integers between. means the first source domain dataset samples, The value of is an integer between 1 and N. The target domain dataset does not participate in model training and is only used for testing, and the source domain participates in training.

[0049] There is a feature extraction network and a classification network in each client. Let the model set of N feature extraction networks be , the model set of N classification networks is . The global feature extraction network model in the central server is set as , the global classification network model is .

[0050] In order to effectively aggregate and use data on the premise of ensuring the data security of different clients and solv...

Embodiment 2

[0093] The second embodiment of the present disclosure provides a mechanical fault diagnosis system based on federal domain generalization, including:

[0094] A central server and a client, the central server is used to initialize the global model, the central server includes a global feature extraction network and a global classification network, and the central server simultaneously exchanges information with multiple clients; each Each client contains a feature extraction network and a classification network. It is assumed that N clients contain N source domain datasets, and the N+1th client contains target domain datasets.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a mechanical fault diagnosis method and system based on federated domain generalization, and relates to the technical field of fault diagnosis. In a training stage, a central server firstly randomly initializes a global model and sends the global model to all clients; secondly, the client side independently trains the model through the training data set of the client side; and 3, sending all models trained by the clients to a server, and averaging all model parameters in the server to obtain a global model. And 4, the client and the central server cooperatively train the global model. And in a test stage, the server sends the global model to a client containing target domain data to complete fault diagnosis. According to the method, the inherent relation between labels and features of source domain data is utilized, and the training loss and model parameters of different client models are weighted and aggregated in a central server, so that the training of a global fault diagnosis model is completed.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis, in particular to a method and system for diagnosing mechanical faults based on federal domain generalization. Background technique [0002] The mechanical fault data usually comes from equipment with different models, different working conditions or different operating environments. The fault diagnosis model jointly trained with these data has the shortcomings of low accuracy and poor generalization ability to predict new data. Domain generalization and domain adaptation methods in transfer learning address the domain drift problem by aligning the data feature space. In domain generalization and domain adaptation methods, labeled data is usually called source domain data, and unlabeled data to be predicted is called target domain data. Domain adaptation realizes fault prediction of target domain data by aligning the source domain and target domain data in feature space. Different from d...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06F21/62G06N3/04G06N3/08G06N20/00
CPCG06N20/00G06N3/08G06F21/6245G06N3/045G06F18/2431
Inventor 宋艳李沂滨贾磊崔明王代超
Owner SHANDONG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products