Rotating machinery fault diagnosis method, system and device based on attribute synthesis conditional diffusion model and distribution offset calibration, and medium

By using the attribute synthesis conditional diffusion model and distribution offset calibration method, high-quality composite fault samples are generated, which solves the problems of sample scarcity and distribution offset in rotating machinery fault diagnosis and improves the generalization ability and stability of the diagnostic model.

CN122153572APending Publication Date: 2026-06-05NAT UNIV OF DEFENSE TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
NAT UNIV OF DEFENSE TECH
Filing Date
2026-02-09
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing methods for diagnosing rotating machinery faults suffer from low sample quality and severe distribution bias due to the scarcity of composite fault samples and high annotation costs, resulting in insufficient generalization ability across operating conditions.

Method used

A conditional diffusion model based on attribute synthesis and a distribution offset calibration method are adopted. Composite fault samples are generated through semantic decoupling and proportional coding mechanisms in a low-dimensional orthogonal attribute space. A classifier-free conditional diffusion model is introduced, combined with a dual indirect offset calibration strategy, to generate high-quality composite fault samples.

Benefits of technology

It significantly improves the physical rationality and feature fidelity of composite fault samples, and enhances the generalization ability and stability of the diagnostic model in zero-sample and cross-condition scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122153572A_ABST
    Figure CN122153572A_ABST
Patent Text Reader

Abstract

The application relates to a rotating machinery fault diagnosis method, system, device and medium based on an attribute-based conditional diffusion model and distribution offset calibration, which comprises the following steps: performing attribute space dimension reduction on single-class fault features and assigning one-hot encoding for identity identification; according to the contribution proportion in the composite fault combination, a dynamic weight vector is configured to weight and synthesize corresponding multiple one-hot encodings to generate a composite fault proportion code; the composite fault proportion code is input into a conditional diffusion model as a guide condition to synthesize a composite fault sample in a high-dimensional feature space; the composite fault sample is subjected to double indirect offset calibration of conditional distribution and prior distribution to obtain a calibrated training data set; a preset fault diagnosis model is trained by using the calibrated training data set, and the trained fault diagnosis model is deployed and applied to perform fault diagnosis and identification on the sensing data of a target rotating machinery. The application can realize zero-sample diagnosis of unknown rotating machinery composite faults.
Need to check novelty before this filing date? Find Prior Art