A5008296183- Machine Fault Diagnosis Techniques
- Gear and Bearing Dynamics Analysis
- Ultrasonics and Acoustic Wave Propagation
- Adversarial Robustness in Machine Learning
- Industrial Vision Systems and Defect Detection
- Integrated Circuits and Semiconductor Failure Analysis
- Anomaly Detection Techniques and Applications
- Structural Health Monitoring Techniques
- Advanced Image Processing Techniques
- Image and Video Quality Assessment
- Fault Detection and Control Systems
- Advanced Image Fusion Techniques
Beijing Institute of Technology
2025
Xi'an Jiaotong University
2018-2023
Artificial neural network (ANN) has achieved great success in mechanical fault diagnosis and been widely used. However, traditional ANN is still opaque terms of interpretability, making it difficult for users to understand trust the results. This paper proposes an interpretable provide high-performance credible The proposed mainly generated by unrolling nested iterative soft thresholding algorithm (NISTA) a sparse coding model named NISTA-Net. Therefore, architecture NISTA-Net clear...
In mechanical anomaly detection, algorithms with higher accuracy, such as those based on artificial neural networks, are frequently constructed black boxes, resulting in opaque interpretability architecture and low credibility results. This article proposes an adversarial algorithm unrolling network (AAU-Net) for interpretable detection. AAU-Net is a generative (GAN). Its generator, composed of encoder decoder, mainly produced by sparse coding model, which specially designed feature encoding...
The adoption of vibration signals in sparse representation (SR) modeling is very popular realizing bearing fault diagnosis recent years. However, it still challenging since the feature information signal usually submerged by strong and complex noise. Thus, noise distribution assumption SR model needs to be carefully studied, namely, modeling. In this article, we propose a new under that obeys mixture generalized Gaussian (MoGG) distribution. Also, L p norm (0 ≤p≤1) adopted as regularization...
Synchrosqueezing transform (SST) has been proposed to characterize frequency-modulated signals with slow varying instantaneous frequency (IF). However, it cannot generate highly concentrated TF representations (TFR) for fast IF, and is easily contaminated by noise. In this paper, band (IFB) its width are defined such that a new TFA method called statistic synchrosqueezing (Stat-SST) the IF of noisy in noise-reduced way. Firstly, we define IFB improve property estimator SST reassign...
Digital images are captured by various fixed and mobile cameras, compressed with traditional novel techniques, transmitted through different communication channels, stored in storage devices. Distortions can occur at each stage of the image acquisition, processing, transmission pipeline, resulting loss perceptual information degradation quality. Therefore, quality assessment is becoming increasingly important monitoring ensuring reliability processing systems. And as most widely applicable...
Sparse representation (SR) theory gets great development in recent years for bearing fault diagnosis. Many scholars focus on constructing proper regularization terms, while few of them notice that the noise assumption is also quite important. Because actual engineering signal, does not necessarily obey a single Gaussian distribution, it usually assumed so traditional SR model. Therefore this paper, we propose new model, which fits signal with generalized distribution (0 ≤q ≤2) and assumes...
Bearing fault diagnosis is one of the most important topics in condition-based maintenance and also a challenging problem because heavy noise interference. Sparse representation methods have been proven effective to solve such problem, especially through L1 norm regularization (Laplacian). In this paper, we analyze sparse signal model deeply Bayesian aspect conclude that distribution wavelet coefficients (transforming by TQWT) well modeled hyper-Laplacian. However, prior makes optimization...