TCN-Transformer-based Automatic Anomaly Detection for Industrial Robots

doi:10.3901/JME.260405

Abstract

Abstract: To address the need for automated anomaly detection in industrial robot factory inspections, a novel industrial robot automatic anomaly detection method based on the TCN-Transformer model is proposed. The method designs a TCN-Transformer-based dynamic modeling approach for industrial robots, where real-time joint angles, angular velocities, and angular accelerations are input into the TCN-Transformer model to perform inverse dynamics. This generates adaptive real-time standard joint current signals, which are compared with the actual real-time current signals to measure similarity and enable adaptive anomaly detection. A cloud-edge collaborative industrial robot automatic anomaly detection system is built, enabling automatic information acquisition, real-time data collection, and automated anomaly detection for robots in the production testing area. Through multiple experimental scenarios, including multi-condition robot tests, joint anomaly injection experiments, and system stress testing, the proposed dynamic modeling method is shown to generate adaptive standard data with high accuracy. The anomaly detection method is able to locate faulty joints and maintain a high consistency with the severity of the anomaly. This system provides stable, accurate, and efficient automated anomaly detection for large-scale industrial robot clusters.

Key words: industrial robots, dynamics modeling, Transformer model, anomaly detection, cloud-edge collaboration

CLC Number:

TP242

JIANG Qincheng, TAO Jianfeng, WANG Shijie, WANG Yangyang, LIU Chengliang. TCN-Transformer-based Automatic Anomaly Detection for Industrial Robots[J]. Journal of Mechanical Engineering, 2026, 62(9): 1-13.

Add to citation manager EndNote|Reference Manager|ProCite|BibTeX|RefWorks

URL: http://www.cjmenet.com.cn/EN/10.3901/JME.260405

http://www.cjmenet.com.cn/EN/Y2026/V62/I9/1

References

[1] YAMADA Y，NAGAMATSU S，SATO Y. Development of multi-arm robots for automobile assembly[C]// Proceedings of 1995 IEEE International Conference on Robotics and Automation，Nagoya：IEEE，1995：2224-2229.
[2] PARK C，PARKark K，GWEON D. Development of industrial dual arm robot for precision assembly of mechanical parts for automobiles[C]// 2006 SICE-ICASE International Joint Conference，2006：3059-3062.
[3] TAO W，ZHANG M，LIU M，et al. Residual vibration analysis and suppression for SCARA robot arm in semiconductor manufacturing[C]// 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems，2006：5153-5158.
[4] International Federation of Robotics. Executive Summary WR 2019 Industrial Robots[EB/OL]. [2019-09-18]. https：//www.ifr.org/free-downloads/.
[5] ARMSTRONG B，KHATIB O，BURDICK J. The explicit dynamic model and inertial parameters of the PUMA 560 arm[C]// IEEE International Conference on Robotics and Automation. San Francisco，USA，1986：510-518.
[6] TRAVERSARO S，BROSSETTE S，ESCANDE A，et al. Identification of fully physical consistent inertial parameters using optimization on manifolds[C]// 2016 IEEE/RSJ International. Conference on Intelligent Robots and Systems (IROS)，2016：5446-5451.
[7] BINGUL Z，KARAHAN O. Dynamic identification of Staubli RX-60 robot using PSO and LS methods[J]. Expert Systems with Applications，2011，38(4)：4136-4149.
[8] GAUTIER M. Extended Kalman filtering and weighted least squaresdynamic identification of robot[J]. Control Engineering Practice，2001，9(12)：1361-1372.
[9] SWEVERS J，VERDONCK W. Dynamic model identification for industrial robots[J]. IEEE//Control Systems Magazine，2007，27(5)：58-71.
[10] 蔡改贫，刘鑫，罗小燕，等. 采用APSO-LM-BP神经网络的挖掘机器人运动学逆解研究[J]. 机械科学与技术，2020，39(5)：706-713. CAI Gaipin，LIU Xin，LUO Xiaoyan，et al. Inverse kinematics solution of mining robot via APSO-LM-BP neural network[J]. Mechanical Science and Technology，2020，39(5)：706-713.
[11] VIVAS A，POIGNET P，MARQUET F，et al. Experimental dynamic identification of a fully parallel robot[C]// 2003 IEEE International Conference on Robotics and Automation (Cat. No.03CH37422)，2003：3278-3283.
[12] 吴文祥. 多自由度串联机器人关节摩擦分析与低速高精度运动控制[D]. 杭州：浙江大学，2014. WU Wenxiang. Joint friction analysis and low-speed high-precision motion control of multi-DOF serial robots[D]. Hangzhou：Zhejiang University，2014.
[13] PIOGNET P，GAUTIER M. Comparison of weighted least squares and extended Kalman filtering methods for dynamic identification of robots[C]// Proceedings 2000 ICRA. Millennium Conference. IEEE International Conference on Robotics and Automation. Symposia Proceedings (Cat. No.00CH37065)，2000：3622-3627.
[14] 齐湛江. SCARA机器人动力学参数辨识及轨迹跟踪控制方法研究[D]. 兰州：兰州理工大学，2022. QI Zhanjiang. The research on dynamic parameter identification and trajectory tracking control of SCARA robot[D]. Lanzhou：Lanzhou University of Technology，2022.
[15] ZHANG L P，MA S G，LI B，et al． BP networks based trajectory planning and inverse kinematics of a reconfigurable mars rover[C]// Proceedings of Second International Conference on Advances in Neural Networks． Chongqing：Springer，2005.
[16] 程韬波，李晓晓，徐智浩，等. 基于递归神经网络的多机器人智能协同控制[J]. 机电工程技术，2020，49(5)：1-4. CHENG Taobo，LI Xiaoxiao，XU Zhihao，et al. Intelligent cooperative control of multiple manipulators based on recurrent neural network[J]. Mechanical and Electrical Engineering Technology，2020，49(5)：1-4.
[17] 贺继林，戴驹，周烜亦. 凿岩机器人末端定位误差补偿[J]. 工程机械，2020，51(9)：45-50. HE Jilin，DAI Ju，ZHOU Xuanyi. Terminal positioning error compensation of rock drilling robot[J]. Construction Machinery，2020，51(9)：45-50.
[18] 刘楠，李兴国，王收军，等. 基于Transformer的重载液压机械臂逆动力学建模[J]. 天津理工大学学报，2024(1)：1-8. LIU Nan，LI Xingguo，WANG Shoujun，et al. Inverse dynamics modeling of heavy-duty hydraulic manipulator based on Transformer[J]. Journal of Tianjin University of Technology，2024(1)：1-8.
[19] 于鸿洋，李文韬，王昭靖. 一种基于物联网的室内墙面抹灰机器人远程监控系统：中国，CN202011457754.1[P]. 2020-12-11. YU Hongyang，LI Wentao，WANG Zhaojing. A remote monitoring system for indoor wall plastering robots based on the Internet of Things：China，CN202011457754.1[P]. 2020-12-11.
[20] BAI S，KOLTER J Z，KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[EB/OL]. arXiv：1803.01271，(2018-04-19).
[21] VASWANI A，SHAZEER N，PARMAR N，et al. Attention is all you need[EB/OL]. arXiv：1706：03762. (2017-06-12).
[22] 刘明新. 基于双向长短时神经网络的机械臂动力学参数辨识[D]. 武汉：武汉科技大学，2024. LIU Mingxin. Identification of dynamic parameters of manipulator based on bidirectional long-short-time neural network[D]. Wuhan：Wuhan University of Science and Technology，2024.
[23] 欧阳志民. 工业机器人关节状态监测方法研究[D]. 昆明：昆明理工大学，2022. OUYANG Zhimin. Research on condition monitoring method of industrial robot[D]. Kunming：Kunming University of Science and Technology，2022.
[24] QIN C，SHI G，TAO J，et al. Precise cutterhead torque prediction for shield tunneling machines using a novel hybrid deep neural network[J]. Mechanical Systems and Signal Processing，2021，151：107386.
[25] DOMHAN T. How much attention do you need? A granular analysis of neural machine translation architectures[C]// GUREVYCH I，MIYAO Y. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1：Long Papers). Melbourne，Australia：Association for Computational Linguistics，2018：1799-1808.
[26] GHEINI M，REN X，MAY J. Cross-attention is all you need：adapting pretrained transformers for machine translation[M]. arXiv，2021.
[27] LIN H，CHENG X，WU X，et al. CAT：Cross attention in vision transformer[M]. arXiv，2021.
[28] 杨琦，陈智才. 基于EMD和相关系数法的列车滚动轴承故障诊断方法研究[J]. 电力机车与城轨车辆，2018，41(3)：15-17. YANG Qi，CHEN Zhicai. Fault diagnosis method for train rolling bearing based on EMD and CCM[J]. Electric Locomotive and Urban Mass Transit Vehicle，2018，41(3)：15-17.