Wenpeng Ma
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Wenpeng Ma has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Peihao Zhu
Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control, Tianjin University of Technology, Tianjin 300384, China
3 shared publications - Qingchun Zheng
Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control, School of Mechanical Engineering, Tianjin University of Technology, Tianjin 300384, China
3 shared publications - Jieyong Deng
Jiangxi Technical College Of Manufacturing, Nanchang 330095, China
2 shared publications Tianchang Zhang
Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control, School of Mechanical Engineering, Tianjin University of Technology, Tianjin, China
2 shared publicationsWenqian Zhang
Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control, School of Mechanical Engineering, Tianjin University of Technology, Tianjin 300384, People’s Republic of China
2 shared publications- Zhiyong Wei
Tianjin Key Laboratory for Advanced Mechatronic System Design and Intelligent Control, Tianjin University of Technology, Tianjin 300384, China
2 shared publications
Latest Publications
- 3 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 74 (4) , 5554-5563
An Enhanced Multifeature Fusion Method for Rotating Component Fault Diagnosis
Planetary gearboxes are critical components in wind turbine systems, where accurate fault diagnosis is essential for maintaining operational stability. However, current signal demodulation methods rely heavily on expert knowledge for demodulated component selection and feature extraction, often struggling to precisely identify fault types and severity levels. To overcome these limitations, this article proposes a novel demodulated component enhanced learning framework, which incorporates three key innovations: first, a proper rotation component reconstruction strategy guided by instantaneous frequency fluctuations to preserve key fault information; second, a method for constructing a feature vector by using characteristic frequencies derived from Amplitude modulation/frequency modulation models to extract key fault amplitudes; third, a hybrid architecture combining feature and model design, in which the long short term memory-convolutional neural network captures spatial and temporal patterns for fault diagnosis across different platforms. Experiments were conducted on two independent test platforms to demonstrate the framework’s superior performance, achieving 99.64% diagnostic accuracy across four gear fault types in case study 1, and 94.48% diagnostic accuracy across four gear fault types and three severity levels in case study 2. The results highlight the method’s strong generalization capability and its potential for practical application in wind turbine condition monitoring systems.
- 3 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Reliability
- Vol. 74 (4) , 5554-5563
An Enhanced Multifeature Fusion Method for Rotating Component Fault Diagnosis
Planetary gearboxes are critical components in wind turbine systems, where accurate fault diagnosis is essential for maintaining operational stability. However, current signal demodulation methods rely heavily on expert knowledge for demodulated component selection and feature extraction, often struggling to precisely identify fault types and severity levels. To overcome these limitations, this article proposes a novel demodulated component enhanced learning framework, which incorporates three key innovations: first, a proper rotation component reconstruction strategy guided by instantaneous frequency fluctuations to preserve key fault information; second, a method for constructing a feature vector by using characteristic frequencies derived from Amplitude modulation/frequency modulation models to extract key fault amplitudes; third, a hybrid architecture combining feature and model design, in which the long short term memory-convolutional neural network captures spatial and temporal patterns for fault diagnosis across different platforms. Experiments were conducted on two independent test platforms to demonstrate the framework’s superior performance, achieving 99.64% diagnostic accuracy across four gear fault types in case study 1, and 94.48% diagnostic accuracy across four gear fault types and three severity levels in case study 2. The results highlight the method’s strong generalization capability and its potential for practical application in wind turbine condition monitoring systems.
- 14 October 2025
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (10) , 106119
Two-stage multi-feature fusion method based on self-attention mechanism and its application to mechanical fault diagnosis
Mechanical fault diagnosis plays a vital role in ensuring the safety and reliability of industrial equipment. Existing techniques, however, exhibit significant limitations in face of multi-faults and multi-modal signals. This article presents a two-stage multi-feature fusion strategy employing a self-attention mechanism (SAM) to enhance fault diagnosis accuracy and robustness. Employing fast Fourier transform (FFT) and wavelet packet transform (WPT), frequency and time-frequency domain features are extracted in the first stage. Convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM) networks are utilized to capture spatial and temporal information in the second stage. The features are fused complementarily by SAM in the two stages respectively. Experimental results demonstrate an impressive diagnostic accuracy of 99.83% and 99.46% on the public dataset and self-collected dataset, respectively. Extensive comparison experiments verify the superiority of the proposed method compared to some others, validating its effectiveness and application potential in mechanical fault diagnosis.
- 14 October 2025
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (10) , 106119
Two-stage multi-feature fusion method based on self-attention mechanism and its application to mechanical fault diagnosis
Mechanical fault diagnosis plays a vital role in ensuring the safety and reliability of industrial equipment. Existing techniques, however, exhibit significant limitations in face of multi-faults and multi-modal signals. This article presents a two-stage multi-feature fusion strategy employing a self-attention mechanism (SAM) to enhance fault diagnosis accuracy and robustness. Employing fast Fourier transform (FFT) and wavelet packet transform (WPT), frequency and time-frequency domain features are extracted in the first stage. Convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM) networks are utilized to capture spatial and temporal information in the second stage. The features are fused complementarily by SAM in the two stages respectively. Experimental results demonstrate an impressive diagnostic accuracy of 99.83% and 99.46% on the public dataset and self-collected dataset, respectively. Extensive comparison experiments verify the superiority of the proposed method compared to some others, validating its effectiveness and application potential in mechanical fault diagnosis.
- Open Access
- 20 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 18457-18469
Fault Diagnosis of Magnetically Controlled On-Column Circuit Breaker Based on Small Sample Condition
As the pivotal component of power grid supply protection, the fault diagnosis of magnetically controlled on-column circuit breakers is a crucial element in ensuring regional power supply reliability. Addressing challenges such as difficult fault signal acquisition, noise interference, and a limited number of fault samples, this paper introduces a fault diagnosis method for magnetically controlled on-column circuit breakers under conditions of small sample sizes, based on VAE-ACGAN-SDAE. Initially, a Variational Autoencoder (VAE) is employed to extract the latent distribution of genuine samples, which are then integrated with the Auxiliary Classifier Generative Adversarial Network (ACGAN) generator to learn the characteristics of real data. Subsequently, to address the problem of real-world operational data being susceptible to noise, a Stacked Denoising Autoencoder (SDAE) is utilized as the discriminator in the ACGAN framework. This approach not only enhances noise resistance and optimizes the feature centers but also synchronizes training with the VAE, thereby improving the quality of the generated samples and refining the weight bias parameters of the discriminator. The experimental outcomes demonstrate that the generated data can be systematically produced and categorized. Optimal sample classification is achieved when the expansion ratio is set to 3, and the method achieves the highest accuracy of 98.8% with a minimal number of fault samples. Compared to the VAE-GAN-CNN network, this network shows a 3.6% increase in accuracy.
- 20 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 18457-18469
Open AccessFault Diagnosis of Magnetically Controlled On-Column Circuit Breaker Based on Small Sample Condition
As the pivotal component of power grid supply protection, the fault diagnosis of magnetically controlled on-column circuit breakers is a crucial element in ensuring regional power supply reliability. Addressing challenges such as difficult fault signal acquisition, noise interference, and a limited number of fault samples, this paper introduces a fault diagnosis method for magnetically controlled on-column circuit breakers under conditions of small sample sizes, based on VAE-ACGAN-SDAE. Initially, a Variational Autoencoder (VAE) is employed to extract the latent distribution of genuine samples, which are then integrated with the Auxiliary Classifier Generative Adversarial Network (ACGAN) generator to learn the characteristics of real data. Subsequently, to address the problem of real-world operational data being susceptible to noise, a Stacked Denoising Autoencoder (SDAE) is utilized as the discriminator in the ACGAN framework. This approach not only enhances noise resistance and optimizes the feature centers but also synchronizes training with the VAE, thereby improving the quality of the generated samples and refining the weight bias parameters of the discriminator. The experimental outcomes demonstrate that the generated data can be systematically produced and categorized. Optimal sample classification is achieved when the expansion ratio is set to 3, and the method achieves the highest accuracy of 98.8% with a minimal number of fault samples. Compared to the VAE-GAN-CNN network, this network shows a 3.6% increase in accuracy.
- 30 December 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (1) , 016343
DE-RGBD SLAM: enhancing static feature point selection in RGB-D visual SLAM using depth information
Feature point extraction plays a key role in visual simultaneous localization and mapping (SLAM) systems. And it remains a major challenge to accurately select static feature points in a complex dynamic environment. To address this issue, this paper proposes an RGB-D SLAM method, referred to as DE-RGBD SLAM, which optimizes feature selection by integrating depth information and effectively utilizes depth data and multi-view geometric information to achieve localization and navigation for mobile robots in dynamic environments. Firstly, the method analyzes prominent feature regions in the image based on color and depth information captured by an RGB-D camera. It sets adaptive FAST corner detection thresholds according to the grayscale information of these regions while masking other areas. Next, the method obtains in-depth information on the detected feature points in the current frame. It combines their pixel coordinates in the image coordinate system to determine the presence of redundant feature points. Notably, the method can detect some dynamic feature points between consecutive frames. Subsequently, in the camera coordinate system, the method compares the depth information of feature points in the depth image with the epipolar depth estimates derived from the essential matrix to determine whether the features are static and eliminate dynamic feature points. This approach significantly enhances the reliability of static feature points. Finally, the accuracy and robustness of the proposed method are validated through experiments conducted on the public TUM dataset and real-world scenarios compared to state-of-the-art visual SLAM systems.
- 30 December 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (1) , 016343
DE-RGBD SLAM: enhancing static feature point selection in RGB-D visual SLAM using depth information
Feature point extraction plays a key role in visual simultaneous localization and mapping (SLAM) systems. And it remains a major challenge to accurately select static feature points in a complex dynamic environment. To address this issue, this paper proposes an RGB-D SLAM method, referred to as DE-RGBD SLAM, which optimizes feature selection by integrating depth information and effectively utilizes depth data and multi-view geometric information to achieve localization and navigation for mobile robots in dynamic environments. Firstly, the method analyzes prominent feature regions in the image based on color and depth information captured by an RGB-D camera. It sets adaptive FAST corner detection thresholds according to the grayscale information of these regions while masking other areas. Next, the method obtains in-depth information on the detected feature points in the current frame. It combines their pixel coordinates in the image coordinate system to determine the presence of redundant feature points. Notably, the method can detect some dynamic feature points between consecutive frames. Subsequently, in the camera coordinate system, the method compares the depth information of feature points in the depth image with the epipolar depth estimates derived from the essential matrix to determine whether the features are static and eliminate dynamic feature points. This approach significantly enhances the reliability of static feature points. Finally, the accuracy and robustness of the proposed method are validated through experiments conducted on the public TUM dataset and real-world scenarios compared to state-of-the-art visual SLAM systems.
- Open Access
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Path planning of mobile robot in dynamic obstacle avoidance environment based on deep reinforcement learning
In this study, to address the issues faced by mobile robots in complex environments, such as sparse rewards caused by limited effective experience, slow learning efficiency in the early stages of training, as well as poor obstacle avoidance performance in environments with dynamic obstacles, the authors proposed a new path planning algorithm for mobile robots by introducing Intrinsic Curiosity Module (ICM) and Long Short-Term Memory (LSTM) into the Proximal Policy Optimization (PPO) algorithm. ICM provided intrinsic rewards in addition to external rewards, accelerating the initial convergence speed. And the Actor-Critic network was optimized by employing a LSTM-based neural network to enhance the performance of avoiding dynamic obstacles. Then, various simulation experiments were conducted in Gazebo with scenarios featuring both static and dynamic obstacles, and the TurtleBot3 mobile robot was used for experimental verification. The experiments demonstrate that the proposed algorithm significantly accelerates convergence in environments with sparse rewards compared to traditional algorithms, and the robot can find more target points within a single episode, indicating more effective path planning ability. Additionally, it can avoid obstacles in various states. Finally, the effectiveness of the algorithm was validated using the TurtleBot3 physical robot in real-world scenarios. Results show that compared to Deep Deterministic Policy Gradient (DDPG), PPO, LSTM-PPO and ICM-PPO, the success rate of the proposed algorithm in path planning increases by 9%, 7.2%, 4% and 4.8% in the most complex simulation environment and by 20%, 16%, 10% and 12% in the physical environment, respectively.
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Open AccessPath planning of mobile robot in dynamic obstacle avoidance environment based on deep reinforcement learning
In this study, to address the issues faced by mobile robots in complex environments, such as sparse rewards caused by limited effective experience, slow learning efficiency in the early stages of training, as well as poor obstacle avoidance performance in environments with dynamic obstacles, the authors proposed a new path planning algorithm for mobile robots by introducing Intrinsic Curiosity Module (ICM) and Long Short-Term Memory (LSTM) into the Proximal Policy Optimization (PPO) algorithm. ICM provided intrinsic rewards in addition to external rewards, accelerating the initial convergence speed. And the Actor-Critic network was optimized by employing a LSTM-based neural network to enhance the performance of avoiding dynamic obstacles. Then, various simulation experiments were conducted in Gazebo with scenarios featuring both static and dynamic obstacles, and the TurtleBot3 mobile robot was used for experimental verification. The experiments demonstrate that the proposed algorithm significantly accelerates convergence in environments with sparse rewards compared to traditional algorithms, and the robot can find more target points within a single episode, indicating more effective path planning ability. Additionally, it can avoid obstacles in various states. Finally, the effectiveness of the algorithm was validated using the TurtleBot3 physical robot in real-world scenarios. Results show that compared to Deep Deterministic Policy Gradient (DDPG), PPO, LSTM-PPO and ICM-PPO, the success rate of the proposed algorithm in path planning increases by 9%, 7.2%, 4% and 4.8% in the most complex simulation environment and by 20%, 16%, 10% and 12% in the physical environment, respectively.
- Open Access
- 11 July 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (14) , 6054
A New Fuzzy Backstepping Control Based on RBF Neural Network for Vibration Suppression of Flexible Manipulator
Flexible manipulators have been widely used in industrial production. However, due to the poor rigidity of the flexible manipulator, it is easy to generate vibration. This will reduce the working accuracy and service life of the flexible manipulator. It is necessary to suppress vibration during the operation of the flexible manipulator. Based on the energy method and the Hamilton principle, the partial differential equations of the manipulator were established. Secondly, an improved radial basis function (RBF) neural network was combined with the fuzzy backstepping method to identify and suppress random vibration during the operation of the flexible manipulator, and the Lyapunov function and control law were designed. Finally, Simulink was used to build a simulation platform, three different external disturbances were set up, and the effect of vibration suppression was observed through the change curves of the final velocity error and displacement error. Compared with the RBF neural network boundary control method and the RBF neural network inversion method, the simulation results show that the effect of the RBF neural network fuzzy inversion method is better than the previous two control methods, the system convergence is faster, and the equilibrium position error is smaller.
- 11 July 2024
- Published by MDPI AG in Applied Sciences
- Vol. 14 (14) , 6054
Open AccessA New Fuzzy Backstepping Control Based on RBF Neural Network for Vibration Suppression of Flexible Manipulator
Flexible manipulators have been widely used in industrial production. However, due to the poor rigidity of the flexible manipulator, it is easy to generate vibration. This will reduce the working accuracy and service life of the flexible manipulator. It is necessary to suppress vibration during the operation of the flexible manipulator. Based on the energy method and the Hamilton principle, the partial differential equations of the manipulator were established. Secondly, an improved radial basis function (RBF) neural network was combined with the fuzzy backstepping method to identify and suppress random vibration during the operation of the flexible manipulator, and the Lyapunov function and control law were designed. Finally, Simulink was used to build a simulation platform, three different external disturbances were set up, and the effect of vibration suppression was observed through the change curves of the final velocity error and displacement error. Compared with the RBF neural network boundary control method and the RBF neural network inversion method, the simulation results show that the effect of the RBF neural network fuzzy inversion method is better than the previous two control methods, the system convergence is faster, and the equilibrium position error is smaller.
- Open Access
- 20 June 2024
- Published by MDPI AG in Preprints.org
A New Fuzzy Backstepping Control Based on RBF Neural Network for Vibration Suppression of Flexible Manipulator
Flexible manipulators have been widely used in industrial production. However, due to the poor rigidity of the flexible manipulator, it is easy to generate vibration. It will reduce the working accuracy and service life of the flexible manipulator. It is necessary to suppress the vibration during the operation of the flexible manipulator. Based on the energy method and Hamilton principle, the partial differential equations of the flexible manipulator are established. Secondly, an improved radial basis function (RBF) neural network is combined with the fuzzy backstepping method to identify and suppress the random vibration during the operation of the flexible manipulator, and the Lyapunov function and control law are designed. Finally, Simulink was used to build a simulation platform, three different external disturbances were set up, and the effect of vibration suppression was observed through the change curves of the final velocity error and displacement error. Compared with RBF neural network boundary control method and RBF neural network inversion method, the simulation results show that the effect of RBF neural network fuzzy inversion method is better than the previous two control methods, the system convergence is faster, and the equilibrium position error is smaller.
- 20 June 2024
- Published by MDPI AG in Preprints.org
Open AccessA New Fuzzy Backstepping Control Based on RBF Neural Network for Vibration Suppression of Flexible Manipulator
Flexible manipulators have been widely used in industrial production. However, due to the poor rigidity of the flexible manipulator, it is easy to generate vibration. It will reduce the working accuracy and service life of the flexible manipulator. It is necessary to suppress the vibration during the operation of the flexible manipulator. Based on the energy method and Hamilton principle, the partial differential equations of the flexible manipulator are established. Secondly, an improved radial basis function (RBF) neural network is combined with the fuzzy backstepping method to identify and suppress the random vibration during the operation of the flexible manipulator, and the Lyapunov function and control law are designed. Finally, Simulink was used to build a simulation platform, three different external disturbances were set up, and the effect of vibration suppression was observed through the change curves of the final velocity error and displacement error. Compared with RBF neural network boundary control method and RBF neural network inversion method, the simulation results show that the effect of RBF neural network fuzzy inversion method is better than the previous two control methods, the system convergence is faster, and the equilibrium position error is smaller.
- 30 September 2023
- Published by Editura Academiei Române in Proceedings of the Romanian Academy, Series A: Mathematics, Physics, Technical Sciences, Information Science
- Vol. 24 (3) , 237-246
Robotic arm trajectory tracking method based on improved proximal policy optimization
To study the method of trajectory tracking for robotic arms, the traditional tracking method has low accuracy and cannot realize the complex tracking tasks. Compared with traditional methods, deep reinforcement learning is an effective scheme with the advantages of robustness and solving complex problems. This study aims to improve the tracking efficiency of robotic arms based on deep reinforcement learning. Thereby, we propose an approach to improve the proximal policy optimization (Improved-PPO) in this paper, which can be applied to multiple degrees of freedom robotic arms for trajectory tracking. In this study, proximal policy optimization (PPO) and model predictive control (MPC) are integrated to provide an effective algorithm for robotic arm applications. MPC is employed for trajectory prediction to design the controller. Further, the Improved-PPO algorithm is employed for trajectory tracking. The Improved-PPO algorithm is further compared with the asynchronous advantage actor-critic (A3C) and PPO algorithms. The simulation results show that the convergence speed of the Improved-PPO algorithm is increased by 84.3% and 15.4% compared with the A3C and PPO algorithms. This method provides a new research concept for robotic arm trajectory tracking.
- 30 September 2023
- Published by Editura Academiei Române in Proceedings of the Romanian Academy, Series A: Mathematics, Physics, Technical Sciences, Information Science
- Vol. 24 (3) , 237-246
Robotic arm trajectory tracking method based on improved proximal policy optimization
To study the method of trajectory tracking for robotic arms, the traditional tracking method has low accuracy and cannot realize the complex tracking tasks. Compared with traditional methods, deep reinforcement learning is an effective scheme with the advantages of robustness and solving complex problems. This study aims to improve the tracking efficiency of robotic arms based on deep reinforcement learning. Thereby, we propose an approach to improve the proximal policy optimization (Improved-PPO) in this paper, which can be applied to multiple degrees of freedom robotic arms for trajectory tracking. In this study, proximal policy optimization (PPO) and model predictive control (MPC) are integrated to provide an effective algorithm for robotic arm applications. MPC is employed for trajectory prediction to design the controller. Further, the Improved-PPO algorithm is employed for trajectory tracking. The Improved-PPO algorithm is further compared with the asynchronous advantage actor-critic (A3C) and PPO algorithms. The simulation results show that the convergence speed of the Improved-PPO algorithm is increased by 84.3% and 15.4% compared with the A3C and PPO algorithms. This method provides a new research concept for robotic arm trajectory tracking.