一种基于多尺度时频特征融合扩散模型的航空发动机振动信号生成方法及系统

By using a multi-scale time-frequency feature fusion diffusion model and employing time-frequency analysis and denoising diffusion techniques, key fault frequency bands of aero-engine vibration signals are identified and generated. This solves the problem of insufficient fault frequency band perception in existing technologies and improves the fidelity of generated samples and fault diagnosis results.

CN122112934BActive Publication Date: 2026-07-17AECC SICHUAN GAS TURBINE RES INST
View PDF 5 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
AECC SICHUAN GAS TURBINE RES INST
Filing Date
2026-04-30
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing technologies lack fault frequency band perception capabilities in generating vibration signals for aero-engines. The generation process treats the entire frequency band equally, failing to concentrate resources on preserving key fault characteristics. Furthermore, the lack of explicit physical guidance based on fault mechanisms leads to the frequency domain structure of the generated samples deviating from the physical laws of actual fault vibration signals, thus reducing fault discriminability.

Method used

A multi-scale time-frequency feature fusion diffusion model is adopted. The frequency domain response intensity and spectral peak significance distribution are extracted through time-frequency analysis. A fault correlation function is constructed to identify key frequency bands. Conditional injection is performed using a denoising diffusion probability model and a cross-attention mechanism to train a noise prediction neural network and generate vibration signals corresponding to the fault type.

Benefits of technology

It improves the fidelity of generated samples in critical fault frequency bands, effectively avoids the problems of scarce aero-engine fault samples and sample class imbalance, and enhances the support capability for fault diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122112934B_ABST
    Figure CN122112934B_ABST
Patent Text Reader

Abstract

本发明涉及航空发动机故障诊断技术领域，公开了一种基于多尺度时频特征融合扩散模型的航空发动机振动信号生成方法及系统，利用振动信号时频表示提取频域响应强度分布及谱峰显著性分布，构建用于表征频率与故障特征关联程度的故障相关度函数，自适应识别关键故障频带；采用包括噪声预测损失和故障感知频谱加权损失的联合损失函数对噪声预测神经网络进行训练，以对故障关键频带施加更强约束，提高生成样本在故障频段的保真度。本发明能有效规避航空发动机运行故障样本稀缺、类别不平衡及现有扩散生成模型对故障频带感知能力不足的问题。
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network

    CN110595775A

  • Hydraulic system intelligent fault diagnosis method based on diffusion model data enhancement framework

    CN120332289A

  • Vibration signal space-time reconstruction method based on multi-modal condition diffusion model

    CN121502240A

  • Method and device for enhancing operation fault data of hydroelectric generating set

    CN120578956A

  • Interpretable deep feature fusion network-based industrial intelligent predictive maintenance method

    WO2026021130A1