Article
Automation & Control Systems
Kaibo Zhou, Chaoying Yang, Jie Liu, Qi Xu
Summary: This article proposes a dynamic graph-based feature learning method for rotating machinery fault diagnosis. Noisy vibration signals are converted into static graphs, where redundant edges are simplified and edge connections are updated based on the relationship among high-level features. The dynamic input graph, constructed as a new graph representation for noisy samples, effectively improves diagnostic performance under different signal-to-noise ratios.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Automation & Control Systems
Xiaoli Zhao, Jianyong Yao, Wenxiang Deng, Peng Ding, Jichao Zhuang, Zheng Liu
Summary: This article proposes a new algorithm called multiscale deep graph convolutional networks (MS-DGCNs) to alleviate the problem of significant disordered fluctuations in the measured signals of the rotor-bearing system. An intelligent fault diagnosis method based on MS-DGCNs is designed to improve feature representations and accuracy. Experimental results demonstrate the higher accuracy and generalization of the method.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Giuseppe Patane
Summary: Data are represented as graphs in various applications, and our goal is to define novel Fourier-based and graph filters induced by rational polynomials for graph processing. We introduce a spectrum-free approach to efficiently evaluate discrete spectral Fourier-based and wavelet operators. Approximating arbitrary graph filters with rational polynomials provides a more accurate and stable alternative. Our proposed approach has advantages in terms of generality and low computational cost.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Engineering, Electrical & Electronic
Xunshi Yan, Yang Liu, Chen-an Zhang
Summary: This article proposes a novel algorithm, the multiresolution hypergraph neural network, which can discover higher-order complex relationships between samples and mine the structure hidden in the data by establishing and fusing hypergraph structures at multiple resolutions.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Automation & Control Systems
Jiamin Xu, Haobin Ke, Zhiwen Chen, Xinyu Fan, Tao Peng, Chunhua Yang
Summary: In this paper, an oversmoothing relief GCN (OsR-GCN) method is proposed to improve the diagnostic performance of the conventional GCN-based fault diagnosis method. Two association graph construction methods and a weight coefficient are introduced in the OsR-GCN model, and an improved particle swarm optimization algorithm is used to find the optimal weight coefficient.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Zonghan Wu, Shirui Pan, Guodong Long, Jing Jiang, Chengqi Zhang
Summary: AutoGCN is proposed to capture the full spectrum of graph signals and automatically update the bandwidth of graph convolutional filters, overcoming the limitations of existing graph convolutional networks that only work as low-pass filters.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Filippo Maria Bianchi, Daniele Grattarola, Lorenzo Livi, Cesare Alippi
Summary: This paper proposes a novel graph convolutional layer based on the auto-regressive moving average (ARMA) filter, which provides a more flexible frequency response, is more robust to noise, and better captures the global graph structure compared to polynomial filters. Experimental results show significant improvements of the proposed ARMA layer over graph neural networks based on polynomial filters.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Asif Salim, S. Sumitra
Summary: This paper introduces the application of graph convolutional neural networks (GCNN) in graph learning and proposes a new framework for filter design. By designing new filters and testing their performance, it is found that they outperform other methods. Additionally, further optimization of the network and related developments are discussed.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Fangming Guo, Zhongwei Li, Ziqi Xin, Xue Zhu, Leiquan Wang, Jie Zhang
Summary: The proposed method integrates spatial and spectral graphs for hyperspectral image classification, using a dual graph u-nets structure. By constructing two graphs for feature extraction and utilizing a dual GCN to extract spatial and spectral graph features simultaneously, effective features are extracted and fused for classification, demonstrating effectiveness in hyperspectral image classification according to experiments on public datasets.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Geochemistry & Geophysics
Chunyu Pu, Hong Huang, Liuyang Luo
Summary: The paper introduces a novel method called attention mechanism-based dual-path convolutional network (AMDPCN) that combines global information learning model and local feature extraction network, along with a multi-scale attention mechanism to improve hyperspectral image classification performance.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Automation & Control Systems
Wenliao Du, Pengjie Hu, Hongchao Wang, Xiaoyun Gong, Shuangyuan Wang
Summary: In this article, a one-dimension in tandem with 2-D joint convolutional neural network (1D-2D JCNN) is proposed for rotating machinery fault diagnosis. It uses 1-D convolution to obtain multiscale feature vectors, constructs them into 2-D maps, and feeds them into a 2-D convolutional neural network. Experimental results show the excellent classification performance of 1D-2D JCNN in bearing and gear fault diagnosis.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Engineering, Mechanical
Wenyi Liu, Jianbo Yu
Summary: This study proposes a novel deep neural network, DSBNet, for machinery fault diagnosis in non-ideal data scenarios. By embedding deep Bayesian learning into DSBNet, feature learning and enhancement of vibration signals are achieved. Through the generation of multi-scale semi-Bayesian features and the application of an adaptive selector, DSBNet demonstrates superior performance in non-ideal training data scenarios.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Multidisciplinary
Leilei Cao, Yan Gu, Chuanzeng Zhang, Qing-Hua Qin
Summary: This paper presents a meshless Chebyshev collocation method (CCM) for solving the eigenvalue problems of the Helmholtz equation, utilizing Chebyshev polynomials for accurate approximation of eigenfunctions to ensure pseudo-spectral convergence. Two approaches, the generalized eigenvalue approach and the standard eigenvalue approach, are implemented to convert the original eigenvalue problem into the generalized and standard forms, respectively. Five benchmark numerical examples are used to demonstrate the accuracy and efficiency of the proposed CCM.
ENGINEERING ANALYSIS WITH BOUNDARY ELEMENTS
(2021)
Article
Engineering, Multidisciplinary
Xiayang Li, Jinzhi Wang
Summary: This paper investigates a fault-tolerant leaderless consensus problem for general linear multi-agent systems (MASs) under a strongly connected directed communication graph. The influences of the actuator bias faults and loss of actuator effectiveness faults on MASs are considered in this paper. The designed fully distributed adaptive fault-tolerant consensus protocols (FTCPs) can make all agents' states achieve consensus, asymptotically. Using the features of the loss of actuator effectiveness faults and the strongly connected graph, novel positive definite Lyapunov functions are constructed skillfully to analyse the validity of the proposed protocols. Finally, simulation examples are presented to demonstrate the effectiveness of the proposed protocols.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2023)
Article
Multidisciplinary Sciences
Batirkhan Turmetov, Valery Karachik
Summary: The study focuses on the eigenfunctions and eigenvalues of the nonlocal Laplace equation with multiple involution. An explicit form of these functions and values for the unit ball is obtained, and a theorem on the completeness of the eigenfunctions is proven.
Article
Automation & Control Systems
Jiahui Cao, Zhibo Yang, Shaohua Tian, Haoqi Li, Ruochen Jin, Ruqiang Yan, Xuefeng Chen
Summary: This article proposes a biprobes method for blade tip timing (BTT) technology to monitor the vibration of rotating blades. The method utilizes aliasing to identify the natural frequencies of blades and employs the robust Chinese remainder theorem for aliasing suppression. Unlike conventional multirate sampling systems, this method uses varying delay durations to obtain different equivalent sample frequencies. The robustness and effectiveness of the method are verified through numerical and experimental validation.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Engineering, Mechanical
Zuogang Shang, Zhibin Zhao, Ruqiang Yan, Xuefeng Chen
Summary: Machine anomaly detection is the task of detecting abnormal machine conditions using collected monitoring data. Autoencoder (AE) based unsupervised anomaly detection (UAD) has gained increasing attention for mechanical equipment. However, the raw monitoring data may be polluted by abnormal data, and without effective regularization, AE-based methods would overfit these polluted data. To address this issue, the core loss is designed to perform AE-based UAD in a model-agnostic and end-to-end manner under data pollution.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Industrial
Fujin Wang, Zhibin Zhao, Zhi Zhai, Zuogang Shang, Ruqiang Yan, Xuefeng Chen
Summary: Deep neural networks have been successfully applied in battery health management for state-of-health (SOH) estimation and remaining useful life (RUL) prediction. However, the lack of transparency in traditional neural networks due to their black-box nature has hindered their explainability. To address this issue, a framework incorporating explanation techniques and model feedback is proposed to improve the performance and explainability of the model for lithium-ion battery SOH estimation. The framework is validated on various datasets and models, demonstrating its superiority in both explaining the model's decisions and enhancing its performance. The code for this framework is open source and available at: https://github.com/wang-fujin/Explainability-driven_SOH.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Instruments & Instrumentation
Hang Lu, Jiawen Xu, Ruqiang Yan
Summary: This research introduces an acoustic wireless energy transmission system that is highly efficient and robust. The system consists of a transmitter and receiver based on a piezoelectric cantilever, coupled through the forces of permanent magnets. By leveraging the strong coupling effect of magnet force, mechanical energy can be wirelessly transferred through the air and metal plate mediums. Experimental studies demonstrate voltage transmission efficiencies of 55.59% and 51.58% for energy transfer through air and air-metal-air mediums, respectively. Furthermore, the system achieves a maximum power transmission of 42.73 mW at an operational frequency of 104.2 Hz. This wireless energy transmission system has applications in enclosed, electrically shielded, and biomedical areas.
REVIEW OF SCIENTIFIC INSTRUMENTS
(2023)
Article
Engineering, Multidisciplinary
Xinxian Chen, Xiaotian Shi, Jiafeng Tang, Zhibin Zhao, Yanjie Guo, Lei Yang
Summary: The paper proposes a fast trend filtering method, called Fast RobustETF, to efficiently and accurately extract the friction coefficient under unknown noise in material friction performance test. It combines mix of Gaussian and non-convex sparsity-inducing functions to extend l1 trend filtering. The method segments the data using a sliding window strategy, distinguishes the wear stage, and selects the characteristic data segment for extraction. It then uses RobustETF in each window and integrates the trend signals to complete the extraction of the friction coefficient.
Article
Engineering, Mechanical
Ruqiang Yan, Zuogang Shang, Hong Xu, Jingcheng Wen, Zhibin Zhao, Xuefeng Chen, Robert X. Gao
Summary: This paper provides a comprehensive overview of the advancement in wavelet transform-based fault diagnosis research over the last decade. It highlights the applications of wavelet transform in traditional fault diagnosis and intelligent fault diagnosis as well as discusses future research trends.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Mechanical
Qisheng Wang, Xin Lin, Xianyin Duan, Ruqiang Yan, Jerry Ying Hsi Fuh, Kunpeng Zhu
Summary: Laser powder bed fusion (L-PBF) is a metal additive manufacturing process with potential for high-performance metal components. However, stability and repeatability issues limit its industrial application. To ensure product quality, process monitoring and control are crucial. A new motion feature is extracted and a classification model is constructed to identify the changing melt pool states during the L-PBF process. The Gaussian process classification (GPC) model achieves better recognition results compared to other models, with an overall recognition rate of 87.1%.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Electrical & Electronic
Xi Chen, Hui Wang, Siliang Lu, Ruqiang Yan
Summary: This article proposes a new method for bearing remaining useful life (RUL) prediction, called RUL-FLTNP, based on federated learning (FL) and Taylor-expansion network pruning. Through collaborative training between a central server and multiple clients, trimmed models are aggregated using the federated averaging (FedAvg) algorithm, offering a promising solution for prognostic problems in data privacy-preserving scenarios.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Automation & Control Systems
Tianfu Li, Chuang Sun, Olga Fink, Yuangui Yang, Xuefeng Chen, Ruqiang Yan
Summary: This article proposes a filter-informed spectral graph wavelet network (SGWN) for intelligent fault diagnosis. SGWN utilizes the spectral graph wavelet convolutional (SGWConv) layer to simultaneously extract low-pass and band-pass features, thereby preventing the over-smoothing problem caused by long-range low-pass filtering. Experimental results on collected solenoid valve dataset and aero-engine intershaft bearing dataset demonstrate that SGWN outperforms comparative methods in terms of diagnostic accuracy and prevention of over-smoothing, and its extracted features are interpretable with domain knowledge.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Engineering, Electrical & Electronic
Shaowen Chen, Shibin Wang, Botao An, Ruqiang Yan, Xuefeng Chen
Summary: In this paper, a new TF analysis method called statistic synchrosqueezing transform (Stat-SST) is proposed to characterize the fast varying IF of noisy signals in a concentrated and noise-reduced way. Stat-SST improves the performance of the IF estimator of SST by defining the instantaneous frequency band (IFB) and its width, and enhances the concentration of TFR by redistributing TF coefficients. By using a threshold obtained by IFB width, Stat-SST distinguishes signal from noise and greatly reduces noise. The validation with both numerical simulation data and practical aero-engine data shows that Stat-SST is superior, more concentrated, and more robust to some existing TFA methods, especially when analyzing signals with fast varying IF.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Botao An, Shibin Wang, Fuhua Qin, Zhibin Zhao, Ruqiang Yan, Xuefeng Chen
Summary: In this article, an adversarial algorithm unrolling network (AAU-Net) is proposed for interpretable mechanical anomaly detection. AAU-Net, a generative adversarial network (GAN), has a mechanism-driven and interpretable network architecture. A multiscale feature visualization approach is introduced to verify the meaningfulness of the encoded features, enhancing user trust in the detection results.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Engineering, Mechanical
Zuogang Shang, Zhibin Zhao, Ruqiang Yan
Summary: Deep learning is becoming popular for fault diagnosis, but the lack of explainability and the extraction of weak fault features from noisy signals are the major challenges. To address these limitations, we propose the SPINN framework that embeds signal processing knowledge into deep learning models. We also develop a denoising fault-aware wavelet network (DFAWNet) as a practical implementation of SPINN, which significantly improves diagnostic performance compared to other explainable or denoising deep learning networks.
CHINESE JOURNAL OF MECHANICAL ENGINEERING
(2023)