Water turbine interpretable fault detection method based on non-intrusive measurement

Through non-invasive measurement combined with wavelet nuclear convolutional autoencoder and Gaussian hybrid hidden Markov model, the low-dimensional timing characteristics of the turbine are extracted, solving the problem of turbine fault detection under complex variable operating conditions, and achieving efficient and capable Explained fault detection effect.

CN119179991AActive Publication Date: 2024-12-24ZHEJIANG UNIV OF TECH

Patent Information

Application Number
CN202411688112.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-25
Publication Date
2024-12-24
Estimated Expiration
2044-11-25

AI Technical Summary

Technical Problem

Existing turbine fault detection methods are difficult to effectively detect faults under complex variable operating conditions, especially based on signal processing methods, it is difficult to extract fault characteristics, deep learning methods lack interpretability, and invasive measurements have problems with equipment damage and stability.

Method used

Using a non-invasive measurement method, the wavelet kernel convolutional autoencoder and Gaussian hybrid hidden Markov model are used to extract low-dimensional timing dependency features through compressing and estimating networks, and jointly optimized loss function is constructed to enhance the model's ability. Explanation and generalization ability.

Benefits of technology

The generalization ability and applicability of fault detection under variable working conditions are achieved, equipment damage is avoided, and the interpretability and detection accuracy of the model are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119179991A_ABST
    Figure CN119179991A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of rotating machinery fault detection, and discloses a water turbine interpretable fault detection method based on non-intrusive measurement, which comprises the following steps: obtaining low-dimensional time sequence dependent feature representation and reconstruction features by using a compression network based on a wavelet kernel convolution auto-encoder; obtaining a reconstruction input sample according to the reconstruction characteristics, and calculating a sample reconstruction error; respectively predicting a mixed membership matrix and a state transition probability matrix by using an independent estimation network, estimating parameters of a Gaussian mixture hidden Markov model according to predicted values, and calculating energy represented by low-dimensional time sequence dependency features; and constructing a joint optimization loss function based on the sample reconstruction error, the energy represented by the low-dimensional time sequence dependency feature and a penalty function, and updating parameters of the compression network and the estimation network based on the joint optimization loss function for training. According to the method, the interpretability of the model is enhanced, and the generalization ability and applicability of fault detection under variable working conditions are improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Fault prediction method based on wavelet neural network and hidden Markov model

    CN110288046A

  • 5G communication protocol anomaly detection method based on hidden semi-Markov model

    CN116016298A

  • Non-intrusive digital-analog linkage control system fault detection method

    CN118311941A

  • Method of recognizing patterns based on markov chain hidden conditional random field model

    US20130124438A1

Cited By

  • Rotating machinery multi-scale feature fault diagnosis method based on non-intrusive signal

    CN119357786A

  • Electromechanical equipment abnormity identification and fault prediction method based on multi-mode OOD detection

    CN120705671A

  • Abnormality identification and fault prediction method of electromechanical equipment based on multimodal OOD detection

    CN120705671B

  • Motor thermal management fault diagnosis method and system

    CN120744651A