Zhenguo Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zhenguo Sun has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
Qiang Chen
Tsinghua University
6 shared publications- Qiang Chen
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China
5 shared publications Wenzeng Zhang
Key Laboratory for Advanced Materials Processing Technology of Ministry of Education, Department of Mechanical Engineering, Tsinghua University, Beijing, China
4 shared publicationsTianzhi Huang
Department of Mechanical Engineering, Tsinghua University, Beijing, China
3 shared publications- Wenzeng Zhang
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China
3 shared publications - Cheng Zou
Department of Mechanical Engineering, Tsinghua University, Beijing 100084, China
2 shared publications
Latest Publications
- 1 July 2024
- Published by Cambridge University Press (CUP) in Journal of Navigation
- Vol. 77 (4) , 469-484
An autonomous localisation method of wall-climbing robots on large steel components with IMU and fixed RGB-D camera
Wall-climbing robots work on large steel components with magnets, which limits the use of wireless sensors and magnetometers. This study aims to propose a novel autonomous localisation method (RGBD-IMU-AL) with an inertial measurement unit and a fixed RGB-D camera to improve the localisation performance of wall-climbing robots. The method contains five modules: calibration, tracking, three-dimensional (3D) reconstruction, location and attitude estimation. The calibration module is used to obtain the initial attitude angle. The tracking and 3D reconstruction module are used jointly to obtain the rough position and normal vector of the robot chassis. For the location module, a normal vector projection method is established to screen out the top point on the robot shell. An extended Kalman filter (EKF) is used to estimate the heading angle in the attitude estimation module. Experimental results show that the positioning error is within 0⋅02 m, and the positioning performance is better than that of the MS3D method. The heading angle error remains within 3⋅1°. The obtained results prove its applicability for the autonomous localisation in low-texture and magnetically disturbed environments.
- 1 July 2024
- Published by Cambridge University Press (CUP) in Journal of Navigation
- Vol. 77 (4) , 469-484
An autonomous localisation method of wall-climbing robots on large steel components with IMU and fixed RGB-D camera
Wall-climbing robots work on large steel components with magnets, which limits the use of wireless sensors and magnetometers. This study aims to propose a novel autonomous localisation method (RGBD-IMU-AL) with an inertial measurement unit and a fixed RGB-D camera to improve the localisation performance of wall-climbing robots. The method contains five modules: calibration, tracking, three-dimensional (3D) reconstruction, location and attitude estimation. The calibration module is used to obtain the initial attitude angle. The tracking and 3D reconstruction module are used jointly to obtain the rough position and normal vector of the robot chassis. For the location module, a normal vector projection method is established to screen out the top point on the robot shell. An extended Kalman filter (EKF) is used to estimate the heading angle in the attitude estimation module. Experimental results show that the positioning error is within 0⋅02 m, and the positioning performance is better than that of the MS3D method. The heading angle error remains within 3⋅1°. The obtained results prove its applicability for the autonomous localisation in low-texture and magnetically disturbed environments.
- Open Access
- 15 December 2022
- Published by IOS Press
- p. 292-303
Localization of Wall Climbing Robot on Cylinder-Shaped Steel
Due to the limitation of sensor application in special environment such as relatively closed and magnetic interference environment, the positioning and the heading angle errors of wall climbing robots accumulate with time. This paper proposes a difference projection localization method based on an external RGB-D camera and a robot-carried inertial measurement unit (IMU). We differential the depth image to obtain the distance change due to the robot occupancy. Then, the 3D point cloud information is converted into 2D image information by projecting the above distances along the normal vector of the robot chassis, which greatly speeds up the computational speed. The position of the robot is calculated by studying the statistical characteristics of the projection. Two EKFs are designed to estimate the attitude, taking the gravity vector and the normal vector of the robot chassis as observation. The experimental results show that the localization error of the wall climbing robot is within 0.017m, and the heading angle error of the attitude estimation is within 3.1∘. The obtained results prove its applicability in self-localization of the wall climbing robot.
- 15 December 2022
- Published by IOS Press
- p. 292-303
Open AccessLocalization of Wall Climbing Robot on Cylinder-Shaped Steel
Due to the limitation of sensor application in special environment such as relatively closed and magnetic interference environment, the positioning and the heading angle errors of wall climbing robots accumulate with time. This paper proposes a difference projection localization method based on an external RGB-D camera and a robot-carried inertial measurement unit (IMU). We differential the depth image to obtain the distance change due to the robot occupancy. Then, the 3D point cloud information is converted into 2D image information by projecting the above distances along the normal vector of the robot chassis, which greatly speeds up the computational speed. The position of the robot is calculated by studying the statistical characteristics of the projection. Two EKFs are designed to estimate the attitude, taking the gravity vector and the normal vector of the robot chassis as observation. The experimental results show that the localization error of the wall climbing robot is within 0.017m, and the heading angle error of the attitude estimation is within 3.1∘. The obtained results prove its applicability in self-localization of the wall climbing robot.
- Open Access
- 15 December 2022
- Published by IOS Press
- p. 314-321
A Multi-Target Detection Method on Distribution Cabinet Based on Improved Faster R-CNN
It is an inevitable trend to detect and recognize the states of different component on distribution cabinet panel more effectively and accurately by using inspection robots instead of manpower. Aiming at the problems of multiple recognition targets and large size difference in image recognition of distribution cabinet panel, an improved Faster R-CNN multi-target detection method is designed. Automatically detect and recognize of different targets on required cabinet panel is achieved by this method. In this paper, Resnet50 is used instead of VGG16 as the feature extraction network in Faster R-CNN, Adam Optimizer is used instead of Momentum, and anchor box size is changed to adapt target size difference on cabinet panel. Experiments show that the improved model has higher accuracy and less computational cost on the required distribution cabinet panel dataset.
- 15 December 2022
- Published by IOS Press
- p. 314-321
Open AccessA Multi-Target Detection Method on Distribution Cabinet Based on Improved Faster R-CNN
It is an inevitable trend to detect and recognize the states of different component on distribution cabinet panel more effectively and accurately by using inspection robots instead of manpower. Aiming at the problems of multiple recognition targets and large size difference in image recognition of distribution cabinet panel, an improved Faster R-CNN multi-target detection method is designed. Automatically detect and recognize of different targets on required cabinet panel is achieved by this method. In this paper, Resnet50 is used instead of VGG16 as the feature extraction network in Faster R-CNN, Adam Optimizer is used instead of Momentum, and anchor box size is changed to adapt target size difference on cabinet panel. Experiments show that the improved model has higher accuracy and less computational cost on the required distribution cabinet panel dataset.
- Open Access
- 1 March 2021
- Published by SAGE Publications in International Journal of Advanced Robotic Systems
- Vol. 18 (2)
A reconfigurable soft wall-climbing robot actuated by electromagnet
This article demonstrates a reconfigurable soft wall-climbing robot actuated by electromagnet. The robot follows the earthworm movement gait and is capable of translation, deflection, and rotation movement while working on a sloping ferromagnetic wall. Also the electromagnetic actuator provides a significant improvement in expeditiousness compared with existing actuation modes. The speed of the robot can be adjusted by modulating the power frequency. When the period of motion cycle is 30 ms, the speed is about 26.5 mm s −1 , and the robot can rotate with a velocity of 14.1° s −1 on the horizontal plane. It can also climb a vertical wall at the speed of 12.6 mm s −1 . The robot is composed of two kinds of modules which can be connected by the magnets embedded. It can also be reconfigured in different working conditions, such as crossing an inaccessible gap, and thus has the potential to be used in flaw detection, surface cleaning, and exploration of ferromagnetic structures.
- 1 March 2021
- Published by SAGE Publications in International Journal of Advanced Robotic Systems
- Vol. 18 (2)
Open AccessA reconfigurable soft wall-climbing robot actuated by electromagnet
This article demonstrates a reconfigurable soft wall-climbing robot actuated by electromagnet. The robot follows the earthworm movement gait and is capable of translation, deflection, and rotation movement while working on a sloping ferromagnetic wall. Also the electromagnetic actuator provides a significant improvement in expeditiousness compared with existing actuation modes. The speed of the robot can be adjusted by modulating the power frequency. When the period of motion cycle is 30 ms, the speed is about 26.5 mm s −1 , and the robot can rotate with a velocity of 14.1° s −1 on the horizontal plane. It can also climb a vertical wall at the speed of 12.6 mm s −1 . The robot is composed of two kinds of modules which can be connected by the magnets embedded. It can also be reconfigured in different working conditions, such as crossing an inaccessible gap, and thus has the potential to be used in flaw detection, surface cleaning, and exploration of ferromagnetic structures.
- 2 August 2019
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 123-134
Design of a Master-Slave Composite Wall Climbing Robot System for Penstock Assembly Welding
- 2 August 2019
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 123-134
Design of a Master-Slave Composite Wall Climbing Robot System for Penstock Assembly Welding
- 1 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1376-1380
A Novel Method Improving the Accuracy of Attitude Estimation of All-position Welding Robot
In recent years there have been excellent results in attitude estimation techniques, which aim to offer precise posture information for various industrial robots. Various tasks especially the inspections and welding on the eroded surface of large-scale hydraulic turbine blade are performed by mobile robots. In those case, the robot moves on curved surfaces along the predetermined route and it is necessary to know its real-time attitude angle for the welding task. In this paper, we propose a novel approach for improving the accuracy of attitude estimation of all-position welding robot using low-cost MEMS inertial sensors. The gyro drift error caused by temperature or bias instability is estimated when the welding robot is at zero velocity in the whole course of navigation. The performance of the proposed approach will be analyzed through a series of experiments by the all-position welding robot developed in our laboratory. To show the performance of the presented method, the experimental results obtained by a low-cost Inertial measurement unit (IMU) and a high-accuracy IMU were compared.
- 1 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1376-1380
A Novel Method Improving the Accuracy of Attitude Estimation of All-position Welding Robot
In recent years there have been excellent results in attitude estimation techniques, which aim to offer precise posture information for various industrial robots. Various tasks especially the inspections and welding on the eroded surface of large-scale hydraulic turbine blade are performed by mobile robots. In those case, the robot moves on curved surfaces along the predetermined route and it is necessary to know its real-time attitude angle for the welding task. In this paper, we propose a novel approach for improving the accuracy of attitude estimation of all-position welding robot using low-cost MEMS inertial sensors. The gyro drift error caused by temperature or bias instability is estimated when the welding robot is at zero velocity in the whole course of navigation. The performance of the proposed approach will be analyzed through a series of experiments by the all-position welding robot developed in our laboratory. To show the performance of the presented method, the experimental results obtained by a low-cost Inertial measurement unit (IMU) and a high-accuracy IMU were compared.
- Open Access
- 22 February 2019
- Published by EDP Sciences in MATEC Web of Conferences
- Vol. 269, 02013
Non-contacted Permanent Magnetic Absorbed Wall-climbing Robot for Ultrasonic Weld Inspection of Spherical Tank
A novel wall-climbing robot for the onsite weld inspecting of spherical tank with Time of Flight Diffraction (TOFD) method has been developed, in order to liberate operators from dangerous and heavy working environment. Patented non-contacted permanent magnetic absorbed technology was adopted to realize reliable and flexible all-position moving along the weld seam on the surface of spherical tank. To ensure stable coupling for TOFD test, a surface adaptive probe holder which can provide constant contact force, have been particularly designed. Equipped with a visual sensing based weld seam tracking unit and industrial PC station, the robot could perform automatic flaw detection remotely even in the darkness environment. Onsite ultrasonic weld inspections have been carried out on a 4000m3spherical tank with 28mm in thickness. It is verified that the robot could accomplish tasks in any position and the acquired TOFD images satisfy the requirements of engineering evaluation.
- 22 February 2019
- Published by EDP Sciences in MATEC Web of Conferences
- Vol. 269, 02013
Open AccessNon-contacted Permanent Magnetic Absorbed Wall-climbing Robot for Ultrasonic Weld Inspection of Spherical Tank
A novel wall-climbing robot for the onsite weld inspecting of spherical tank with Time of Flight Diffraction (TOFD) method has been developed, in order to liberate operators from dangerous and heavy working environment. Patented non-contacted permanent magnetic absorbed technology was adopted to realize reliable and flexible all-position moving along the weld seam on the surface of spherical tank. To ensure stable coupling for TOFD test, a surface adaptive probe holder which can provide constant contact force, have been particularly designed. Equipped with a visual sensing based weld seam tracking unit and industrial PC station, the robot could perform automatic flaw detection remotely even in the darkness environment. Onsite ultrasonic weld inspections have been carried out on a 4000m3spherical tank with 28mm in thickness. It is verified that the robot could accomplish tasks in any position and the acquired TOFD images satisfy the requirements of engineering evaluation.
- 11 August 2017
- Published by Springer Nature in Metallurgical and Materials Transactions A
- Vol. 48 (10) , 4622-4631
The Effects of Borides on the Mechanical Properties of TLPB Repaired Inconel 738 Superalloy
- 11 August 2017
- Published by Springer Nature in Metallurgical and Materials Transactions A
- Vol. 48 (10) , 4622-4631
The Effects of Borides on the Mechanical Properties of TLPB Repaired Inconel 738 Superalloy
- Open Access
A Fast Method to Calculate the Spatial Impulse Response for 1-D Linear Ultrasonic Phased Array Transducers
A method is developed to accurately determine the spatial impulse response at the specifically discretized observation points in the radiated field of 1-D linear ultrasonic phased array transducers with great efficiency. In contrast, the previously adopted solutions only optimize the calculation procedure for a single rectangular transducer and required approximation considerations or nonlinear calculation. In this research, an algorithm that follows an alternative approach to expedite the calculation of the spatial impulse response of a rectangular linear array is presented. The key assumption for this algorithm is that the transducer apertures are identical and linearly distributed on an infinite rigid plane baffled with the same pitch. Two points in the observation field, which have the same position relative to two transducer apertures, share the same spatial impulse response that contributed from corresponding transducer, respectively. The observation field is discretized specifically to meet the relationship of equality. The analytical expressions of the proposed algorithm, based on the specific selection of the observation points, are derived to remove redundant calculations. In order to measure the proposed methodology, the simulation results obtained from the proposed method and the classical summation method are compared. The outcomes demonstrate that the proposed strategy can speed up the calculation procedure since it accelerates the speed-up ratio which relies upon the number of discrete points and the number of the array transducers. This development will be valuable in the development of advanced and faster linear ultrasonic phased array systems.
Open AccessA Fast Method to Calculate the Spatial Impulse Response for 1-D Linear Ultrasonic Phased Array Transducers
A method is developed to accurately determine the spatial impulse response at the specifically discretized observation points in the radiated field of 1-D linear ultrasonic phased array transducers with great efficiency. In contrast, the previously adopted solutions only optimize the calculation procedure for a single rectangular transducer and required approximation considerations or nonlinear calculation. In this research, an algorithm that follows an alternative approach to expedite the calculation of the spatial impulse response of a rectangular linear array is presented. The key assumption for this algorithm is that the transducer apertures are identical and linearly distributed on an infinite rigid plane baffled with the same pitch. Two points in the observation field, which have the same position relative to two transducer apertures, share the same spatial impulse response that contributed from corresponding transducer, respectively. The observation field is discretized specifically to meet the relationship of equality. The analytical expressions of the proposed algorithm, based on the specific selection of the observation points, are derived to remove redundant calculations. In order to measure the proposed methodology, the simulation results obtained from the proposed method and the classical summation method are compared. The outcomes demonstrate that the proposed strategy can speed up the calculation procedure since it accelerates the speed-up ratio which relies upon the number of discrete points and the number of the array transducers. This development will be valuable in the development of advanced and faster linear ultrasonic phased array systems.
- Open Access
Mode Conversion Behavior of Guided Wave in a Pipe Inspection System Based on a Long Waveguide
To make clear the mode conversion behavior of S0-mode lamb wave and SH0-plate wave converting to the longitudinal mode guided wave and torsional mode guided wave in a pipe, respectively, the experiments were performed based on a previous built pipe inspection system. The pipe was wound with an L-shaped plate or a T-shaped plate as the waveguide, and the S0-wave and SH0-wave were excited separately in the waveguide. To carry out the objective, a meander-line coil electromagnetic acoustic transducer (EMAT) for S0-wave and a periodic permanent magnet (PPM) EMAT for SH0-wave were developed and optimized. Then, several comparison experiments were conducted to compare the efficiency of mode conversion. Experimental results showed that the T(0,1) mode, L(0,1) mode, and L(0,2) mode guided waves can be successfully detected when converted from the S0-wave or SH0-wave with different shaped waveguides. It can also be inferred that the S0-wave has a better ability to convert to the T(0,1) mode, while the SH0-wave is easier to convert to the L(0,1) mode and L(0,2) mode, and the L-shaped waveguide has a better efficiency than T-shaped waveguide.
Open AccessMode Conversion Behavior of Guided Wave in a Pipe Inspection System Based on a Long Waveguide
To make clear the mode conversion behavior of S0-mode lamb wave and SH0-plate wave converting to the longitudinal mode guided wave and torsional mode guided wave in a pipe, respectively, the experiments were performed based on a previous built pipe inspection system. The pipe was wound with an L-shaped plate or a T-shaped plate as the waveguide, and the S0-wave and SH0-wave were excited separately in the waveguide. To carry out the objective, a meander-line coil electromagnetic acoustic transducer (EMAT) for S0-wave and a periodic permanent magnet (PPM) EMAT for SH0-wave were developed and optimized. Then, several comparison experiments were conducted to compare the efficiency of mode conversion. Experimental results showed that the T(0,1) mode, L(0,1) mode, and L(0,2) mode guided waves can be successfully detected when converted from the S0-wave or SH0-wave with different shaped waveguides. It can also be inferred that the S0-wave has a better ability to convert to the T(0,1) mode, while the SH0-wave is easier to convert to the L(0,1) mode and L(0,2) mode, and the L-shaped waveguide has a better efficiency than T-shaped waveguide.