Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Industrial and Systems Engineering | en_US |
| dc.contributor.advisor | Zhang, Xiaoge (ISE) | en_US |
| dc.creator | Liao, Jingxiao | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/14387 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Signal processing-empowered deep learning for prognostics and health management of rolling bearings | en_US |
| dcterms.abstract | In recent years, deep learning has achieved significant success in various fields, including natural language processing, autonomous driving, and computer vision. In the realm of prognostics and health management (PHM) for rolling bearings in rotating machinery—such as aero engines, wind turbines, and high-speed trains—numerous intelligent PHM methodologies have emerged to provide accurate and adaptable machinery fault diagnostics and prognostics. However, methodologically speaking, there is no one-size-fits-all approach. It is widely acknowledged that these data-driven approaches still have considerable limitations, hindering their widespread adoption in industrial settings. | en_US |
| dcterms.abstract | Three primary challenges persist: (1) the lack of interpretability in deep learning methods, particularly in machinery fault diagnosis, where diagnostic models must be transparent to foster trust in the results and inform maintenance decisions; (2) the limited generalizability and reliability of bearing remaining useful life (RUL) prediction models. When training data is scarce, even under identical operating conditions and with the same bearing types, current RUL models demonstrate suboptimal accuracy. In addition, ensuring the reliability of RUL predictions is an important consideration for making informed maintenance decisions in real-world scenarios; and (3) the difficulty in deploying intelligent diagnosis models to edge devices, which hinders their integration into real-world industrial settings. | en_US |
| dcterms.abstract | Therefore, this dissertation aims to address these challenges by establishing a new paradigm of integrating traditional signal processing and modern deep learning methods. We formally define this approach as signal processing-empowered neural networks, which synthesize the complementary strengths of both domains. This framework provides three key advantages: (1) integrating rigorous signal processing theory to improve model interpretability; (2) leveraging the robust feature representation capabilities of signal processing techniques to enhance deep learning model generalizability and auxiliary exponential model to quantify the reliability of RUL predictions; and (3) enabling faster computation and greater accuracy, thereby facilitating the edge device deployment of lightweight models. The research contents are summarized as follows: | en_US |
| dcterms.abstract | (1) To address the interpretability issue in bearing fault diagnosis, we introduce the concept of neural blind deconvolution (Neural BD), which neuralizes the conventional blind deconvolution (BD) approach. We develop a classifier-guided blind deconvolution (ClassBD) framework that jointly learns BD extracted features and deep learning-based fault labels. The physics guaranteed features can be extracted by BD to provide interpretable information for intelligent fault diagnosis models. | en_US |
| dcterms.abstract | Moreover, the superiority of quadratic neural networks in Neural BD inspired us to enhance the feature extraction ability for unsupervised anomaly detection. We mathematically prove that a heterogeneous network composed of quadratic and conventional neurons can approximate a function with a polynomial number of neurons, whereas a homogeneous network requires an exponential number of neurons to achieve the same level of approximation error. This theoretical result motivates the development of heterogeneous autoencoders (An autoencoder integrates conventional and quadratic neurons). Our proposed heterogeneous autoencoder exhibits both efficiency and effectiveness in unsupervised bearing fault anomaly detection. | en_US |
| dcterms.abstract | (2) To address the generalizability issue of bearing remaining useful life (RUL) prediction, we introduce a logarithmic cumulative transformation (LCT) operator, consisting of cumulative, logarithmic, and another cumulative transformation, for feature extraction. This operator effectively aligns the diverse patterns across vibration degradation data of different bearings. The LCT can be seamlessly embedded into any deep learning or machine learning backbones to enhance the performance of RUL prediction. | en_US |
| dcterms.abstract | Additionally, for reliable RUL estimation, we propose a fast method to quantify the reliability of each prediction by integrating a linear regression model and an auxiliary exponential model. This method provides an assurance on the reliability of model predictions. | en_US |
| dcterms.abstract | (3) To improve the deployability of bearing fault diagnosis models, we propose a lightweight and deployable model for bearing fault diagnosis, referred to as BearingPGA-Net, to address the edge hardware deployment issue. Based on decoupled knowledge distillation, we construct a signal processing and deep learning integrated one-convolutional-layer bearing fault diagnosis model. This lightweight model is successfully deployed in an industrial-level FPGA, facilitating online bearing fault diagnosis with potential for real-world applications. | en_US |
| dcterms.extent | xxii, 197 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2026 | en_US |
| dcterms.educationalLevel | Ph.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.accessRights | open access | en_US |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/14387