Article
Engineering, Mechanical
Yonghao Miao, Chenhui Li, Huifang Shi, Te Han
Summary: This paper proposes a novel deep network-based maximum correlated kurtosis deconvolution (MCKD-DeNet) method to solve the problems of traditional deconvolution methods using neural network and correlated kurtosis. The proposed method initializes the filter using the Hanning window and learns the fault feature with different filters. Correlated kurtosis is used as the cost function to train the neural network, and the input period is estimated by calculating the autocorrelation of the most informative filtered signal. Finally, the component with the most fault information is selected as the output of MCKD-DeNet based on the correlation coefficient.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Industrial
Ruxue Bai, Zong Meng, Quansheng Xu, Fengjie Fan
Summary: In this paper, a novel data representation method based on fractional Fourier transform (FRFT) and recurrence plot transform is proposed for machinery fault diagnosis. Experimental results show that the proposed method outperforms conventional methods such as Fourier spectrum and short time Fourier transform. The fusion of maximum kurtosis based fractional Fourier domain recurrence plot and time domain recurrence plot achieves the best performance, making the trained convolutional neural network adaptive to variable working conditions.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Engineering, Mechanical
Zechao Liu, Shaopu Yang, Yongqiang Liu, Jianhui Lin, Xiaohui Gu
Summary: The performance of the wheelset-bearing system is crucial for the running safety, stability and comfort of high-speed trains. The proposed adaptive correlated Kurtogram (ACK) method can effectively identify different cyclo-stationary components in the vibration signals of wheelset-bearing systems. By adaptively generating the paving of the plane for Kurtogram and highlighting special periodic impulses, ACK improves the effectiveness of identifying frequency bands containing impact information and is capable of detecting multiple frequency bands simultaneously.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Physics, Multidisciplinary
Shan Wang, Pingjuan Niu, Zijian Qiao, Yongfeng Guo, Fuzhong Wang, Chenghao Xu, Shuzhen Han, Yan Wang
Summary: This study introduces an adaptive unsaturated stochastic resonance method using maximum cross-correlated kurtosis as a signal detection index to locate fault signals, combining cross correlated coefficient and spectrum kurtosis. The method is validated with actual vibration signals from motor and gear bearings, showing it is more suitable for explaining periodic impulse components compared to other denoising methods.
CHINESE JOURNAL OF PHYSICS
(2021)
Article
Chemistry, Multidisciplinary
Weihan Li, Yang Li, Ling Yu, Jian Ma, Lei Zhu, Lingfeng Li, Huayue Chen, Wu Deng
Summary: The KMVMD-PGMCKD method integrates the advantages of KMVMD and PGMCKD to construct a novel weak fault feature extraction model, which effectively extracts the fault features of bearing rolling elements and accurately diagnoses weak faults under variable working conditions.
APPLIED SCIENCES-BASEL
(2021)
Article
Engineering, Multidisciplinary
Yating Hou, Xingcheng Han, Jiansheng Bai, Liming Wang
Summary: In this paper, the adaptive scale chirplet transform (ASCT) is proposed to address the issues of biased estimation of instantaneous frequency (IF) and poor noise immunity in current time-frequency (TF) analysis methods. The core idea of the algorithm is to use a frequency-dependent quadratic polynomial kernel function to approximate the IF and a time-varying window length to overcome the frequency resolution problem caused by signal modulation. The ASCT algorithm shows high TF aggregation and effective noise suppression in experimental results, demonstrating its practical relevance.
MEASUREMENT SCIENCE AND TECHNOLOGY
(2023)
Article
Instruments & Instrumentation
Yuanyuan Sheng, Huanyu Liu, Lu Li, Junbao Li
Summary: This work proposes a hybrid model combining frequency-weighted energy operator (FWEO) with power spectrum fusion (PSF) to identify weak fault features of bearings and detect different fault types. The model effectively reduces noise interference through PSF denoising and enhances cyclic fault signals using FWEO. The presence of faults is identified by observing the squared envelope spectrum. Experimental results show that the proposed model has high anti-noise performance and robustness, and can extract fault frequencies well.
REVIEW OF SCIENTIFIC INSTRUMENTS
(2023)
Article
Engineering, Mechanical
Wanyang Zhang, Taihuan Wu, Baoqiang Zhang, Huageng Luo
Summary: In this paper, a new nonlinear basis time-frequency analysis method called multiple squeezing based on velocity synchronous chirplet transform (MSVSCT) is proposed. The MSVSCT utilizes the second-order velocity-synchronous chirplet transform to generate a nonlinear basis that complies with the variations in instantaneous frequency (IF), allowing for the separation of synchronous components even in situations with closely spaced frequencies. The multi-squeezing technique is employed to enhance the energy concentration and accuracy of the time-frequency representation (TFR). The analysis results demonstrate that MSVSCT achieves higher accuracy, robustness to noise, observability, and fidelity in signal reconstruction compared to other popular TFA methods when dealing with synchronous multi-component signals.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Acoustics
Zexing Ni, Dan He, Xiufeng Wang, Ying Zhang
Summary: This paper proposes a vehicle axle abnormal sound detection method based on abnormal sound mechanism, which determines the sound quality of newly produced axles through subjective evaluation and deconvolutes the axle vibration signal using autocorrelation kurtosis. The impulse autocorrelation kurtosis index is constructed to quantify the abnormal sound of the axle. Experimental results show that this index has a correlation of more than 0.9 with the subjective evaluation results, which can better identify the sound quality of the axle.
Article
Engineering, Electrical & Electronic
Quansheng Xu, Jingbo Liu, Yang Guan, Dengyun Sun, Zong Meng
Summary: This article proposes a new time-frequency analysis method for analyzing rolling bearing fault signals, which can effectively represent its instantaneous frequency trajectory and has excellent performance in TF location and noise robustness.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Electrical & Electronic
Xiang Lu, Ao Zhu, Yaqi Song, Guoli Ma, Xingzhen Bai, Yinjing Guo
Summary: This paper proposes a novel hybrid fault diagnosis method for rolling element bearings, which combines complementary complete ensemble robust local mean decomposition with adaptive noise and maximum correlated kurtosis deconvolution based on multiple disturbance multi-verse optimizer. The method effectively improves the fault diagnosis accuracy of rolling element bearings.
Article
Computer Science, Information Systems
LianHui Jia, LiJie Jiang, YongLiang Wen, Hongchao Wang
Summary: A two-stage feature extraction method for early weak fault of rolling element bearing (REB) is proposed, which combines feature mode decomposition (FMD) with blind deconvolution (BD) method to solve the difficulty of fault feature extraction in the early weak fault stage caused by strong background noise interference of REB. The complex original vibration signal is decomposed into single-component modes using FMD, and the mode containing sensitive fault feature is selected for further analysis. The impulsive characteristic is enhanced by applying the deconvolution method, and satisfactory fault features are extracted through traditional envelope spectrum (ES) analysis on the filtered signal.
INTERNATIONAL JOURNAL OF DISTRIBUTED SENSOR NETWORKS
(2023)
Article
Computer Science, Information Systems
Zichang Liu, Siyu Li, Rongcai Wang, Xisheng Jia
Summary: This article proposes a method for extracting faint faults in rolling bearings based on SSA, VMD, and MCKD, with a success rate of up to 100%. By adaptively determining the parameters in VMD and MCKD, it enhances the fault impact components in the signals and effectively extracts the fault characteristic frequencies of rolling bearings.
Article
Chemistry, Analytical
Shishuai Wu, Jun Zhou, Tao Liu
Summary: This paper proposes a fault feature extraction approach combining adaptive variational modal decomposition (AVMD) and improved multiverse optimization (IMVO) algorithm parameterized maximum correlated kurtosis deconvolution (MCKD), which can efficiently extract the acoustic signal fault features of compound faults in rolling bearings.
Article
Engineering, Electrical & Electronic
Qiuyang Zhou, Yuhui Zhang, Jiayin Tang, Jianhui Lin, Liu He, Cai Yi
Summary: The article proposes a new deconvolution method, MICG-Lp/Lq-D, based on the ICG-Lp/Lq norm, which is designed to have robustness to strong noise and sensitivity to repetitive impulses simultaneously. Verification through simulated and experimental case studies shows that the method overcomes the defects of MCG-Lp/Lq-D and can be considered a promising tool for detecting repetitive impulses.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Automation & Control Systems
Hongwei Wang, Jiawei Xiang, Xufeng Zhao, Yulong Li
Summary: A precision reliability evaluation model considering all time-variant and random error sources is proposed in this study. Time-variant wear models for commonly occurred degradation types are developed, and inserted into the precision model with all moving components using the meta-action structural decomposition method. The precision reliability of rotor-bearing systems is solved using the time-variant stochastic process discretization method. Case investigations confirm the performance of the present model.
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2023)
Article
Engineering, Electrical & Electronic
Qiang Gao, Jun Young Jeon, Jiawei Xiang, Gyuhae Park
Summary: This article presents a non-equidistant T-shaped sensor cluster for localizing the impact on structures. The normal T-shaped sensor cluster configuration effectively removes blind areas by combining equidistant linear sensor array beamforming and an L-shaped sensor cluster method. However, the equidistant linear sensor array beamforming suffers from spatial aliasing effect, resulting in low localization accuracy. To address this, the proposed non-equidistant T-shaped sensor cluster utilizes nonequidistant linear sensor array beamforming method to improve impact localization accuracy in the regions near vertical directions.
IEEE SENSORS JOURNAL
(2023)
Article
Engineering, Electrical & Electronic
Yizhu Shi, Yuqing Zhou, Yan Ren, Weifang Sun, Jiawei Xiang
Summary: This research proposes a hybrid method for identifying the spring energy storage state in circuit breaker operating mechanisms. The method utilizes the Gramian angular field (GAF) to represent the dynamic characteristics evolution process and combines it with a convolutional block attention module (CBAM) and residual network (ResNet). Experimental results demonstrate the effectiveness of the proposed method.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Engineering, Industrial
Yi Liu, Hang Xiang, Zhansi Jiang, Jiawei Xiang
Summary: Intelligent fault diagnosis methods are effective in ensuring the safety and reliability of key parts of rotating machinery. However, the lack of data during equipment acceptance period and the assumption of high data quality affect the reliability of results. To address these issues, a time-frequency-based method is introduced to analyze impulse components based on fault features. An accurate time-frequency analysis method named the second-order transient-extracting S-transform is proposed to overcome the influence of uncertain parameters. It produces a highly concentrated time-frequency representation and demonstrates higher accuracy in feature detection compared to other methods.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Chemistry, Analytical
Qizhe Lin, Xiaoqi Li, Bicheng Tu, Junwei Cao, Ming Zhang, Jiawei Xiang
Summary: This study proposes a two-stage method for estimating the SOC of a lithium-ion battery by combining a second-order resistor-capacitor (RC) equivalent circuit model and an eXogenous Kalman filter (XKF). The SOC estimation values obtained by a stable observer are fed into XKF to enhance accuracy and stability. Experimental results show that the proposed method outperforms the commonly used EKF method in SOC estimation.
Article
Engineering, Marine
Hu Jiang, Yongying Jiang, Jiawei Xiang
Summary: Structural damage detection based on vibration signals has gained attention for its easy implementation. However, this method is not directly applicable to structures in operation. To address this limitation, a two-stage multi-damage detection method using operating deflection shape (ODS) and random forest (RF) techniques has been developed. It utilizes 2D wavelet transform and finite element method (FEM) to identify damage locations and establish relationships between natural frequencies and damage severities. Numerical simulations and experimental tests show that this method achieves acceptable damage-detection precision.
Article
Chemistry, Analytical
Yi Liu, Hang Xiang, Zhansi Jiang, Jiawei Xiang
Summary: Intelligent fault diagnosis of roller bearings faces challenges from similar distribution of train and test datasets and limitations in installation positions of accelerometer sensors in industrial environments. This paper proposes a DA-ResNet model using MMD and a residual connection for cross-domain diagnosis of roller bearings based on acoustic and vibration data. Experimental cases verify the necessity of multi-source data and improvement in recognition accuracy through transfer learning.
Article
Automation & Control Systems
Junkang Zheng, Hui Wang, Anil Kumar, Jiawei Xiang
Summary: The use of machine data-driven artificial intelligence models is important for diagnosing faults in rotary vector (RV) reducers, but the lack of complete fault sample labeled data poses a challenge. To overcome this, a lumped parameter model of an RV reducer is developed to generate sufficient training samples for the AI models.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Electrical & Electronic
Jiahao Li, Yi Liu, Jiawei Xiang
Summary: Maximum CYCBD is a blind deconvolution method used to extract faults in mechanical systems. It faces challenges of setting the cyclic frequency in advance and defining a suitable filter length. To address these issues, an optimal maximum CYCBD method is developed. It involves processing raw signals, estimating the cyclic frequency, and determining the filter length adaptively. The proposed method is validated with data from a bearing dataset and experimental test rigs.
IEEE SENSORS JOURNAL
(2023)
Article
Engineering, Electrical & Electronic
Xiao Zhuang, Di Zhou, Jing Cai, Hongfu Zuo, Xufeng Zhao, Jiawei Xiang
Summary: In this study, a novel image processing and deep normalized convolutional neural network (DNCNN) for location measurement and reading distance prediction of RFID multitags are proposed. An experimental platform is designed involving image technology to measure the reading distance and collect the images of RFID multitags. The proposed DNCNN effectively predicts the reading distance of RFID multitags by deeply mining the relationship between the tags' location distributions and corresponding reading distance.
IEEE SENSORS JOURNAL
(2023)
Article
Engineering, Multidisciplinary
Qi Wei, Jiawei Xiang, Weiping Zhu, Hongjiu Hu
Summary: A novel method combining WBEM with frequency-independent fundamental solutions is proposed to determine the band structures of fluid-solid PCs. Integral equations are based on frequency-independent fundamental solutions, avoiding nonlinear eigenvalue problems and reducing computing time. Domain integral terms are handled with RIM and DRM. B-spline wavelet and wavelet coefficients are used to approximate the physical boundary conditions. Coupling conditions and Bloch theorem are applied to wavelet coefficients. Matrix compression technique is used to truncate small matrix entries generated by wavelet vanishing moment characteristics. Examples are provided to demonstrate the accuracy and efficiency of the method.
INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Yuqing Zhou, Hongche Wang, Gonghai Wang, Anil Kumar, Weifang Sun, Jiawei Xiang
Summary: In recent years, deep learning-based methods have made remarkable achievements in the intelligent fault diagnosis of rotating machinery. However, the lack of labeled and large unlabeled samples in actual industrial scenes affects the performance of supervised learning methods. This paper proposes a novel semi-supervised fault diagnosis method based on multiscale permutation entropy (MPE) enhanced contrastive learning (CL). Experimental results in gearbox and milling tool fault diagnosis experiments show that the proposed MPE-CL method outperforms other benchmark methods with classification accuracy above 95.4% and 96.0% when the labeled training dataset size is 50/class, respectively.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Engineering, Electrical & Electronic
Yi Liu, Hang Xiang, Zhansi Jiang, Jiawei Xiang
Summary: An improved general linear chirplet transform method is developed to overcome the limitations of traditional short-time Fourier transform (STFT)-based methods in processing non-stationary signals. By iteratively upgrading the instantaneous frequency (IF) and introducing a synchrosqueezing operator, the method improves the estimation accuracy of IF curves and enhances the concentration of time-frequency representation under variable operating conditions. Experiments with simulated data confirm the effectiveness of the enhanced time-frequency analysis (TFA) method, showing superior sharpness of IF curves and enhanced time-frequency readability compared to other advanced TFA methods, as well as superior feature extraction ability.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Chemistry, Analytical
Hanlin Guan, Ren Yan, Hesheng Tang, Jiawei Xiang
Summary: This paper proposes an effective intelligent fault diagnosis method for hydraulic multi-way valves using an improved Squeeze-Excitation Convolution Neural Network and Gated Recurrent Unit (SECNN-GRU). The method extracts statistical and fault features, and combines convolutional neural network and recurrent neural network for feature extraction and fusion, achieving high accuracy fault diagnosis.
Article
Engineering, Industrial
Hongwei Wang, Yaqi Liu, Zongyi Mu, Jiawei Xiang, Jian Li
Summary: This paper proposes a hybrid approach supported by digital twins for real-time precision reliability prediction of worm drive systems. It establishes a virtual mirror and uses a Wiener process-based degradation model to handle the data and predict precision reliability. The accuracy and effectiveness of the proposed method are verified by comparing the predicted results driven by different methods.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)