FINAL PROGRAM and BOOK OF ABSTRACTSFINAL PROGRAM and BOOK OF ABSTRACTS 2020 IEEE 9th Data Driven Control and Learning Systems Conference (DDCLS’20) Liuzhou, China November 20–22,

FINAL PROGRAM and BOOK OF ABSTRACTS

2020 IEEE 9th Data Driven Control and Learning

Systems Conference (DDCLS’20)

Liuzhou, China

November 20–22, 2020

Organized by Technical Committee on Data Driven Control, Learning and Optimization, Chinese Association of Automation

Beijing Jiaotong University Qingdao University

Locally Organized by Guangxi University of Science and Technology

Sponsored by IEEE Beijing Section

IEEE Industrial Electronics Society IEEE CIS Beijing Chapter

Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the Publisher. IEEE Catalog Number: CFP20HAG-USB

ISBN: 978-1-7281-5921-8

CONTENTS Organizing Committee………………………………………………………………………………...1

Welcome Message from General Chairs……………………………………………………………3

Message from Technical Program Chairs…………………………………………………………...5

Keynote Address………………………………………………………………………………………7

Distinguished Lecture ……………………………………………………………………………… 11

Technical Program and Book of Abstracts…………………………………………………………21

Program at a Glance…………………………………………………………………………………86

DDCLS’20

1

Organizing Committee

General Chairs: Zhongsheng Hou, Qingdao University, China Simin Li, Guangxi University of Science and Technology, China General Co-Chairs: Chenghong Wang, Chinese Association of Automation, China Xiongxiong He, Zhejiang University of Technology, China Guangxing Tan, Guangxi University of Science and Technology, China Organizing Committee Chairs: Jing Wang, Beijing University of Chemical Technology, China Xisheng Dai, Guangxi University of Science and Technology, China Technical Program Committee Chairs: Mingxuan Sun, Zhejiang University of Technology, China Huaguang Zhang, Northeastern University, China Regional Chairs: Xiao’e Ruan, Xi’an Jiaotong University, China Junmin Li, Xidian University, China Fei Liu, Jiangnan University, China Yong Fang, Shanghai Universtiy, China Zhiqiang Ge, Zhengjiang University, China Xiaodong Li, Sun Yat-sen University, China Xiangyang Li, South China University of Technology, China Li Wang, North China University of Technology, China Tianjiang Hu, Sun Yat-sen University, China Aihua Zhang, Bohai University, China Yanjun Liu, Liaoning University of Technology, China Deqing Huang, Southwest Jiaotong University, China Haisheng Yu, Qingdao University, China Ying Zheng, Huazhong University of Technology, China Ruizhuo Song, University of Science & Technology Beijing, China Wenchao Xue, Academy of Mathematics and Systems Science, China

Academy of Sciences, China Chuansheng Wang, Qingdao University of Science & Technology, China Committee Members: Members of Technical Committee on Data Driven Control, Learning and

Optimization and Invited Experts Invited Session Chairs: Zengqiang Chen, Nankai University, China Darong Huang, Chongqing Jiaotong University, China Jing Na, Kunming University of Science and Technology, China Fei Qiao, Tongji University, China Senping Tian, South China University of Technology, China Qinglai Wei, Institute of Automation, Chinese Academy of Sciences, China Zhanshan Wang, Northeastern University, China Jinpeng Yu, Qingdao University, China Weiwei Che, Qingdao University, China Yi Liu, Zhejiang University of Technology, China Jiayan Wen, Guangxi University of Science and Technology, China Subject Session Chairs: Zhihuan Song, Zhejiang University, China Dongbin Zhao, Institute of Automation, Chinese Academy of Sciences, China Xin Xu, National University of Defense Technology, China Panel Discussion Chairs: Hongye Su, Zhejiang University, China Qunxiong Zhu, Beijing University of Chemical Technology, China

2

Zengguang Hou, Institute of Automation, Chinese Academy of Sciences, China Changhua Hu, Rocket Force University of Engineering, China Zhijian Ji, Qingdao University, China Poster Session Chairs Xuhui Bu, Henan Polytechnic University, China Wei Ai, South China University of Technology, China Hongtao Ye , Guangxi University of Science and Technology, China International Affairs Chairs: Danwei Wang, Nanyang Technological University, Singapore Chiang-Ju Chien, Huafan University, Taiwan, China Zhi-Qiang Gao, Cleveland State University, USA Youqing Wang, Shandong University of Science and Technology, China Shen Yin, Harbin Institute of Technology, China Bin Chu, University of Southampton, UK Finance Chairs: Shangtai Jin, Beijing Jiaotong University, China Rongmin Cao, Beijing Information Science and Technology University, China Publication Chairs: Mengqi Zhou, IEEE Beijing Section, China Dong Shen, Renmin University of China, China Editorial Chairs: Ronghu Chi, Qingdao University of Science & Technology, China Yongchun Fang, Nankai University, China Shan Liu, Zhejiang University, China Deyuan Meng, Beihang University, China Publicity Chairs: Weisheng Chen, Xidian University, China Long Cheng, Institute of Automation, Chinese Academy of Sciences, China Shuguo Yang, Qingdao University of Science & Technology, China Liang Cai, Guangxi University of Science and Technology, China Secretaries: Chenkun Yin, Beijing Jiaotong University, China Shoufeng Zhang, Guangxi University of Science and Technology, China Yuwei Zhang, Guangxi University of Science and Technology, China Xiangsuo Fan, Guangxi University of Science and Technology, China

DDCLS’20

3

Welcome Message from General Chairs

Zhongsheng Hou General Chair of DDCLS’20

Simin Li General Chair of DDCLS’20

Dear Friends and Colleagues, On behalf of the Organizing Committee, it is our greatest pleasure to welcome you to the 2020 IEEE 9th Data Driven Control and Learning Systems Conference (DDCLS’20), which is organized by Technical Committee on Data Driven Control, Learning and Optimization (DDCLO), Chinese Association of Automation, Beijing Jiaotong University and Qingdao University, locally organized by Guangxi University of Science and Technology, all are from China, and sponsored by IEEE Beijing Section, IEEE Industrial Electronics Society, IEEE CIS Beijing Chapter. The conference is held at Liuzhou Liudong Ramada Plaza Hotel, Liuzhou, Guangxi Province, China, November 20–22, 2020. Data driven control and learning systems, together with model-based control methods forming the complete control theory, is an emerging hot research area in the field of automation engineering and in systems & control community. It focuses on control, learning and optimization for the plants whose models are unavailable. Although the study on data driven control and learning is still in the embryonic stage, it has attracted a great amount of attention within the systems and control community, such as the special issues published in the top journals: ACTA AUTOMATICA SINICA (2009), IEEE Transactions on Neural Networks (2011), Information Sciences (2013), IEEE Transactions on Industrial Informatics (2013), IEEE Transactions on Industrial Electronics (2015, 2017), and IET Control Theory & Applications (2015, 2016). The keyword ‘Data Driven Control’ was formally listed with the application code F030110 as a new research domain in the project catalog of the National Natural Science Foundation of China in 2019. Further, the data driven control and learning systems would be fundamental challenges in the coming age of the Internet of Things, Cyber-Physical Systems, Industry 4.0, China Manufacturing 2025, and Artificial Intelligence 2.0 under the big data environment, which is already on our road ahead but beyond the traditional systems & control methods.

4

As an inheritance of previous seven workshops, DDCLS’20 continues to attract broad interest throughout the world, with the submission of 317 papers. This reflects the increasing interest in our field, and meanwhile creates a difficult workload in evaluating the papers and organizing a cohesive program. We are fortunate to have an exceptional Technical Program Committee (TPC) that sorted through the evaluations and integrated the individual submissions into the final technical program described in the proceedings. We also want to thank our Organizing Committee for their invaluable assistance in arranging the diverse offerings at the conference, from registration and local arrangements to technical programs. Last but not least, we would like to express our deep appreciation to Guangxi University of Science and Technology for their great support. The Technical Program Committee has assembled a comprehensive technical program that covers a broad spectrum of topics in data driven control and learning systems. The DDCLS’20 technical program comprises 14 regular sessions, 16 invited sessions, 1 best paper award session and 2 interactive sessions. Besides the technical sessions, the highlights of the DDCLS’20 are the keynote addresses given by distinguished senior scholars including Prof. Frank Allgöwer from Germany, Prof. Alessandro Astolfi from UK, Prof. Ben M. Chen from Hong Kong, China and Prof. Derong Liu from China, and the distinguished lectures given by active young scholars including Prof. Chunhui Zhao, Prof. Xiaoli Luan, Prof. Jun Zhao, Prof. Xiaosheng Si, Prof. Yuanjing Feng, Prof. Dong Shen, Prof. Keyou You and Prof. Qiuye Sun, all from China. We sincerely appreciate all the contributors, keynote address speakers, distinguished lecture speakers, invited session organizers, and session chairs for their tremendous efforts towards a top-quality conference. We also want to thank the young lovely volunteers who have made this conference possible. Without you, the monumental task ahead of us for organizing this conference would be significantly beyond our capabilities. May you have a wonderful and fascinating stay in Liuzhou, Guangxi Province, China, and enjoy the colorful scenery and magic foods. Best wishes

Zhongsheng Hou Simin Li

General Chair of DDCLS’20 General Chair of DDCLS’20

DDCLS’20

5

Message from Technical Program Chairs

Mingxuan Sun Technical Program Chair

Huaguang Zhang Technical Program Chair

Dear Friends and Colleagues,

On behalf of the Technical Program Committee, it is our great honor to welcome you to the 2020 IEEE 9th Data Driven Control and Learning Systems Conference (DDCLS’20) in Liuzhou, China.

The annual event of DDCLS has proven to be one of the excellent forums for scientists, researchers, engineers, and industrial practitioners to present and discuss the latest technological advancements as well as future directions and trends in Data Driven Control, Learning and Optimization, and to set up useful links for their works. DDCLS’20 has received enthusiastic responses with a total of 317 submissions. All the submissions had been processed by the Technical Program Committee. All committee members worked professionally, responsibly, and diligently. Besides evaluations from reviewers, each member also provided his/her own assessments on the assigned papers, so as to ensure that only high-quality papers would be accepted. Their commitment and hard work have enabled us to put together a very solid proceeding for our conference. The proceeding includes 261 papers which are divided into 31 oral sessions and 2 poster sessions for presentation.

Ahead of the parallel technical sessions, we will have four keynote talks to be delivered by eminent scientists. These lectures will address the state-of-the-art developments and leading-edge research topics in both theory and applications in Data Driven Control, Learning and Optimization. We are indeed honored to have Prof. Frank Allgöwer (University of Stuttgart), Prof. Alessandro Astolfi (Imperial College London), Prof. Ben M. Chen (Chinese University of Hong Kong), and Prof. Derong Liu (Guangdong University of Technology) as the keynote address speakers. Besides, we are very lucky to have the following distinguished lectures given by eight outstanding young scholars, they are Prof. Chunhui Zhao (Zhejiang University), Prof. Xiaoli Luan (Jiangnan University), Prof. Jun Zhao (Dalian University of Technology), Prof. Xiaosheng Si (Rocket Force University of Engineering), Prof. Yuanjing Feng (Zhejiang University of Technology), Prof. Dong Shen (Renmin University of China), Prof. Keyou You (Tsinghua University) and

6

Prof. Qiuye Sun (Northeastern University). We are confident that their presences would undoubtedly act prestige to the conference. We would like to express our sincere appreciations to all of them for their enthusiastic contributions and strong supports to DDCLS’20. To promote the development of the society of Data Driven Control, Learning and Optimization, the highest quality papers will be rewarded with the Best Paper Award at DDCLS’20. Based on reviewers' comments and nominations as well as the evaluations of Technical Program Committee members, 24 papers were selected for the consideration of the award by the Best Paper Award Committee. These papers were sent to some distinguished experts in the relevant areas for additional evaluations in a double-blind manner. Based on their comments and recommendations, six papers were shortlisted as the finalists for the award. During the conference, the oral presentations of the six finalists will be further assessed by the DDCLS’20 Best Paper Award Committee. The winner of the "DDCLS Best Paper Award" will be selected by the committee after assessing the oral presentations. Furthermore, the interactive presentations of 75 papers in 2 poster sessions will be assessed by the DDCLS’20 Best Poster Award Committee during the conference, and one or two papers will be conferred to the "DDCLS Best Poster Award" by the committee after assessing the interactive presentations.

A U-disk containing the PDF files of all papers scheduled in the program and an Abstract Book will be provided at the conference to each registered participant as part of the registration material. The official conference proceedings will be published by the IEEE and included in the IEEE Xplore Database.

On behalf of the Technical Program Committee, we would like to thank all reviewers for giving time and expertise to provide comments, which are contributive to the Committee in making a fair decision on the acceptance/rejection of each paper. Thanks also go to the dedication, diligence, and commitments of the Invited Session Chairs Prof. Zengqiang Chen, Prof. Darong Huang, Prof. Jing Na, Prof. Fei Qiao, Prof. Senping Tian, Prof. Qinglai Wei, Prof. Zhanshan Wang, Prof. Jinpeng Yu, Prof. Weiwei Che, Prof. Yi Liu, and Prof. Jiayan Wen, Subject Session Chairs Prof. Zhihuan Song, Prof. Dongbin Zhao, Prof. Xin Xu, and all the members of the Technical Program Committee. We would like to gladly acknowledge the technical sponsorship provided by the Organizing Committee of DDCLS’20 and Technical Committee on Data Driven Control, Learning and Optimization, Chinese Association of Automation. We also convey our heartfelt thanks to friends, colleagues, and families who have helped us in completing the technical program directly or indirectly. Last but not least, we are grateful for the strong and enthusiastic support of all delegates, especially those old faces around the world.

We do hope that you will find your participation in DDCLS’20 in Liuzhou is really stimulating, rewarding, enjoyable, and memorable.

Mingxuan Sun Huaguang Zhang

Technical Program Chair Technical Program Chair

DDCLS’20

7

Keynote Address

Keynote Address 1

Reinforcement Learning for Optimal Control Prof. Derong Liu

Guangdong University of Technology, China

Saturday, Nov. 21, 2020 08:30-09:30

Dongcheng Hall / 东城厅

Abstract Reinforcement learning (RL) is one of the most important branches of artificial intelligence.

Researchers have been using RL techniques in modern control theory. Self-learning control methodologies are a good representative of such efforts. RL recently has become a major force in the machine learning fields. On the other hand, adaptive dynamic programming (ADP) has now become popular in control communities. Both RL and ADP have roots in dynamic programming and in many ways they are equivalent. Major breakthroughs of ADPRL for optimal control were achieved around 2006, when iterative ADP approaches were introduced. The optimal control of nonlinear systems requires to solve the nonlinear Bellman equation instead of the Riccati equation as in the linear case. The discrete-time Bellman equation is more difficult to work with than the Riccati equation because it involves solving nonlinear partial difference equations. Though dynamic programming has been a useful computational technique in solving optimal control problems, it is often computationally untenable to run it to obtain the optimal solution, due to the backward numerical process required for its solutions, i.e., the well-known "curse of dimensionality". Self-learning optimal control based on ADPRL provides efficient tools for tackling the following two problems. (1) Nonlinear Bellman equation is solved using iterative ADP approaches which are shown to converge. (2) Neural networks are employed for function approximation in order to obtain forward numerical process. Some new developments in ADPRL for optimal control will be summarized.

Biography Derong Liu received the PhD degree in electrical engineering from the University of Notre Dame in 1994. He became a Full Professor of Electrical and Computer Engineering and of Computer Science at the University of Illinois at Chicago in 2006. He was selected for the “100 Talents Program” by the Chinese Academy of Sciences in 2008, and he served as the Associate Director of The State Key Laboratory of Management and Control for Complex Systems at the Institute of Automation, from 2010 to 2015. He has published 19 books. He is the Editor-in-Chief of Artificial Intelligence Review (Springer). He was the Editor-in-Chief of the IEEE Transactions on Neural Networks and Learning Systems from 2010 to 2105. He is a Fellow of the IEEE, a Fellow of the International Neural

Network Society, and a Fellow of the International Association of Pattern Recognition.

8

Keynote Address 2

Fully Autonomous UAS and Its Applications Prof. Ben M. Chen

Chinese University of Hong Kong, China National University of Singapore, Singapore

Saturday, Nov. 21, 2020 9:30-10:30


Abstract The research and market for the unmanned aerial systems (UAS), or drones, has greatly expanded

over the last few years. It is expected that the currently small civilian unmanned aircraft market is likely to become one of the major technological and economic stories of the modern age, due to a wide variety of possible applications and added value related to this potential technology. Modern unmanned aerial systems are gaining promising success because of their versatility, flexibility, low cost, and minimized risk of operation. In this talk, we highlight some key techniques involved in developing fully autonomous unmanned aerial vehicles and their industrial application examples, which includes deep tunnel inspection, stock counting and checking in warehouses and building inspections. Biography

Ben M. Chen is currently a Professor in the Department of Mechanical and Automation Engineering at the Chinese University of Hong Kong. He was a Provost's Chair Professor in the Department of Electrical and Computer Engineering, the National University of Singapore (NUS), where he was also serving as the Director of Control, Intelligent Systems and Robotics Area, and Head of Control Science Group, NUS Temasek Laboratories. His current research interests are in unmanned systems, robust control and control applications.

Dr. Chen is an IEEE Fellow. He has published more than 400 journal and conference articles, and a dozen research monographs in control theory and applications, unmanned systems and financial market modeling by Springer in New York and London. He had served on the editorial boards of several

international journals including IEEE Transactions on Automatic Control and Automatica. He currently serves as an Editor‐in‐Chief of Unmanned Systems. Dr. Chen has received a number of research awards nationally and internationally. His research team has actively participated in international UAV competitions, and won many championships.

DDCLS’20

9

Keynote Address 3 Data and/or Control – Is Control Theory Becoming Obsolete?

Prof. Frank Allgöwer University of Stuttgart, German

Saturday, Nov. 21, 2020 11:00-12:00


Abstract While recent years have shown rapid progress of learning-based methods to effectively utilize data for

control tasks, most existing control theoretic approaches still require knowledge of an accurate system model. It is worth asking if this trend towards data-driven approaches will ultimately lead to an obsolescence of classical systems and control theory. On the other hand, a key feature of control theory has always been its ability to provide rigorous theoretical guarantees – something that the learning community has only recently begun to address. In this talk, we present a novel framework for data-driven control theory, which does not rely on any model knowledge but still allows to give desirable theoretical guarantees. This framework relies on a result from behavioral systems theory, where it was proven that the vector space of all input-output trajectories of a linear time-invariant system is spanned by time-shifts of a single measured trajectory, given that the respective input signal is persistently exciting. We show how this result can be utilized to develop a mathematically sound approach to data-driven system analysis, with the possibility to verify input-output properties (e.g., dissipation inequalities) of unknown systems. Moreover, we propose a novel purely data-driven model predictive control scheme and we present theoretical results on closed-loop stability and robustness. Finally, the presented framework allows us to design state-feedback controllers with performance guarantees, even if the data are affected by noise. Biography

Frank Allgöwer is director of the Institute for Systems Theory and Automatic Control and professor in Mechanical Engineering at the University of Stuttgart in Germany. Frank's main interests in research and teaching are in the area of systems and control with a current emphasis on the development of new methods for data-based control, optimization-based control, networks of systems, and systems biology. Frank received several recognitions for his work including the IFAC Outstanding Service Award, the IEEE CSS Distinguished Member Award, the State Teaching Award of the German state of Baden-Württemberg, and the Leibniz Prize of the Deutsche Forschungsgemeinschaft.

Frank has been the President of the International Federation of Automatic Control (IFAC) for the years 2017-2020. He was Editor for the journal Automatica from 2001 to 2015 and is editor for the Springer Lecture Notes in Control and Information Science book series and has published over 500 scientific articles. From 2012 until 2020 Frank also served a Vice-President of Germany's most important research funding agency the German Research Foundation (DFG).

10

Keynote Address 4

Data-Driven Model Reduction Prof. Alessandro Astolfi

Imperial College London, UK University of Rome Tor Vergata, Italy

Sunday, Nov. 22, 2020 8:30-9:30

VIP23 Hall / VIP23 厅

Abstract The aim of the talk is to discuss two methods for obtaining reduced order models, for linear and

nonlinear systems, from data. In the first part of the talk the notion of moment for linear systems is generalized to nonlinear, possibly time-delay, systems. It is shown that this notion provides a powerful tool for the identification of reduced order models from input-output data. It is also shown that the canonical parameterization of the reduced order model as a rank-one update of the "interpolation-point matrix" is not necessary, hence one can prove robustness of data-driven model reduction algorithms against variations in the location of the interpolation points. In the second part of the talk the Loewner framework for model reduction is discussed and it is shown that the introduction of left- and right- Loewner matrices/functions simplifies the construction of reduced order models from data.

This is joint work with Z. Wang (Southeast University), G. Scarciotti (Imperial College) and J. Simard (Imperial College). Biography

Alessandro Astolfi was born in Rome, Italy, in 1967. He graduated in electrical engineering from the University of Rome in 1991. In 1992 he joined ETH-Zurich where he obtained a M.Sc. in Information Theory in 1995 and the Ph.D. degree with Medal of Honor in 1995 with a thesis on discontinuous stabilisation of nonholonomic systems. In 1996 he was awarded a Ph.D. from the University of Rome "La Sapienza" for his work on nonlinear robust control. Since 1996 he has been with the Electrical and Electronic Engineering Department of Imperial College London, London (UK), where he is currently Professor of Nonlinear Control Theory and Head of the Control and Power Group. From 1998 to 2003 he was also an Associate Professor at the Dept. of Electronics and Information of the Politecnico of Milano. Since 2005 he has also been a Professor at Dipartimento di Ingegneria Civile e Ingegneria Informatica, University of Rome Tor Vergata. His research interests are

focussed on mathematical control theory and control applications, with special emphasis for the problems of discontinuous stabilisation, robust and adaptive control, observer design and model reduction.

DDCLS’20

11

Distinguished Lecture 1

Data-Driven Wide-Range Nonstationary Process Monitoring Prof. Chunhui Zhao

Zhejiang University, China.

Saturday, Nov. 21, 2020 13:00-14:00


Abstract Modern industrial production often has wide-range nonstationary operating characteristics, such as

batch manufacturing processes, wide-load power generation processes, etc. Due to its large-scale non-stationary operation characteristics, it raises new challenges to the safe and reliable operation of industrial processes and has become the focus of attention. Starting from the traditional batch process, this report will present the concept of a generalized batch process, analyze the specific characteristics of wide-range nonstationary industrial processes, and summarize the basic process monitoring techniques and the relevant research work in this field. It further analyzes the existing specific problems, and extend the traditional batch process analysis methods to industrial processes with wide-range non-stationary operation characteristics. Finally, the application of the proposed method in different fields will be briefly introduced.

Biography Chunhui Zhao has been a Professor with the College of Control Science and Engineering, Zhejiang University, Hangzhou, China. Her research interests include statistical machine learning and data mining for industrial application. She has authored or coauthored more than 120 papers in peer-reviewed international journals. She has published 2 monographs and authorized 21 invention patents. She has hosted more than 10 scientific research projects, including the NSFC funds, provincial projects and corporate cooperation projects. She was the recipient of the National Top 100 Excellent Doctor Thesis Nomination Award, New Century Excellent Talents in University, China, and the National Science Fund for Excellent Young Scholars, respectively. She has also obtained the first Automation Society Young Women Scientist Award, the

Process Control Youth Award, etc., and is now an IEEE senior member. She has served AE of three International Journals, including Journal of Process Control, Control Engineering Practice and Neurocomputing, and two domestic journals, including Control and Decision, and Control Engineering.

12

Distinguished Lecture 2 Distributed Gradient Tracking for Optimization and Learning over Network

Prof. Keyou You Tsinghua University, China

Saturday, Nov. 21, 2020 14:00-14:30


Abstract Many problems of recent interest in control and machine learning can be posed in the framework of

mathematical optimization. As data gets larger and more distributed, distributed algorithms over networks offer ample opportunities to improve the speed and accuracy of optimization. In this talk, we shall exploit the distributed gradient tracking technique (DGT) to solve large-scale optimization and learning problems, e.g., the fully Asynchronous DGT which is easy to implement in directed networks with distributed datasets and robust to bounded transmission delays, while maintaining a linear convergence rate if local functions are strongly-convex with Lipschitz-continuous gradients. Moreover, we adopt the DGT to design distributed algorithms with explicit convergence rates for the distributed resource allocation and distributed training over networks, respectively. Experiments are included to show their advantages against the-state-of-the-art algorithms. Biography

Keyou You received the B.S. degree in Statistical Science from Sun Yat-sen University, Guangzhou, China, in 2007 and the Ph.D. degree in Electrical and Electronic Engineering from Nanyang Technological University (NTU), Singapore, in 2012. After briefly working as a Research Fellow at NTU, he joined Tsinghua University in Beijing, China where he is now an Associate Professor with tenure in the Department of Automation. He held visiting positions at Politecnico di Torino, The Hong Kong University of Science and Technology, The University of Melbourne and etc.

His current research interests include networked control systems, distributed algorithms and learning, and their applications. Dr. You received the Guan Zhaozhi award at the 29th Chinese Control Conference in 2010, a CSC-IBM China Faculty

Award in 2014, and the ACA Temasek Young Educator Award in 2019. He was selected to the National 1000-Youth Talent Program of China in 2014 and received the National Natural Science Fund for Excellent Young Scholars in 2017.

DDCLS’20

13


Prediction and Scheduling for Industrial Energy System Prof. Jun Zhao

Dalian University of Technology, China

Saturday, Nov. 21, 2020 14:30-15:00


Abstract Industrial energy resource saving is capable of not only improving the enterprise profits, but also

carrying out the significant strategy meaning for our country. Given the fixed technical process and equipment, the optimization scheduling of the industrial energy system is the most important approach for such a goal. However, the most industrial energy systems exhibit a very complicated structure, which can hardly establish a mechanism based model to describe such a system, and the existing manual scheduling method makes the decision making process tardily. A class of data-driven predictive scheduling methodology is proposed. In detail, considering the consistent modeling, the quantitative uncertainty description, and the semantic characteristics of the energy data, the short-term prediction model, the prediction interval one and the long-term model are respectively reported, and a rolling optimization technique with the procedures of prediction-scheduling-validation is proposed. The mentioned approaches have been successfully applied to a number of industrial enterprises in our country.

Biography Jun Zhao is now the director of Intelligent Control Institute with the School of Control Science and Engineering, DUT, China. He has authored or co-authored over 100 technical publications in refereed journals and conference proceedings. He serves as associate editors for several top tier journals including Control Engineering Practice, IEEE TNNLS, Information Sciences, etc. From 2015, he became a Technical Committee member (TC6.2) of IFAC MMM society. In 2018, he obtained the First Class Prizes of Science and Technology Progress Award of CAA (Chinese Automation Association), and is now the scientist-in-chief of a National Key R&D Program of China. In addition, he was the recipient of Young Scholar of Yangtze River from Ministry of Education of China in 2016, and received the Excellent Young Scholar funding supported by National Natural Science Foundation of China in 2015. He is also the recipients of the Best Application Paper Award of

WCICA2014, and the Zhang Zhongjun Best Paper Award of CPCC 2016.

DDCLS’20

15


Data Driven Brain Neurofiber Tract Identification Prof. Yuanjing Feng

Zhejiang University of Technology, China

Saturday, Nov. 21, 2020 15:00-15:30


Abstract Accurate brain neurofiber tract identification promises to have a high impact in fundamental

neuroscience and its clinical applications. However, state-of-the-art fiber tracking algorithms are driven by local symmetrically orientation fields estimated from diffusion MRI, representing the local tangent direction to the white matter tract of interest. Conceptually, the principle of inferring connectivity with streamline prorogation from local symmetrically orientation fields can lead to problems as soon as pathways overlap, cross, branch, and have complex geometries. Usually, tractography-based connectome is dominated by lots of false-positive connections. This project will propose an asymmetric tensor stream-flow fiber tracking methods and its application in cranial nerve atlas reconstruction. Firstly, a stream-flow differential equation based on computational fluid mechanics will be presented for describing nerve fiber bundle. The asymmetric fiber geometries is expressed as the distribution of streamline cluster in tensor vector field which extends the streamline prorogation to more general stream-flow way. Then, an data driven automated neurofiber tract identification algorithm will be proposed for connectome-based cranial nerve tractographic atlas based on asymmetric global fiber tracking. Its potential clinical applications in neurosurgical planning and neurodegenerative diseases are presented.

Biography Yuanjing Feng holds a Ph.D in control science and engineering from Xi’an Jiaotong University, M.S. in Mechanical design and theory from Northwest A&F University. Currently, he is the Director of the Institute of Information Processing and Automation and is working as a professor at Zhejiang University of Technology. He is a visiting scholar from January 2010 to February 2012 and October 2018 to April 2019 in Laboratory of Mathematics in Imaging at Harvard University, where he worked with Professor Carl-Fredric Westin. To date, he has authored more than 40 peer-reviewed journal articles (including Automatica, Medical Image Analysis, NeuroImage, Brain research) and MICCAI,

ISBI. His interests include data driven modeling and optimization in field of intelligence transportation system, medical image analysis.

16


Non-Intrusive Modeling for We-Energy based on Mechanism-Data Hybrid Drive

Prof. Qiuye Sun Northeastern University, China

Sunday, Nov. 22, 2020

9:50-10:20 VIP23 Hall / VIP23 厅

Abstract Generally, an accurate model can describe the operating status of a system more effectively and

provide a more reliable theoretical basis for the system optimization and control. To distinguish from the traditional invasive modeling, a non-invasive modeling method based on mechanism and data hybrid is proposed for we-energy, a typical energy system. By using this method, non-invasive modeling for the energy system including photovoltaic, wind power, energy storage devices and energy coupling devices can be carried out. Firstly, the meteorological data, energy output and price curve are utilized to analyze and extract the characteristic of we-energy, and then the characteristic database is established. Afterwards, by taking the port energy data of we-energy as the random noise input, the GAN generator is improved and more applicable to we-energy characteristic. The feedback evaluation of GAN discriminator is utilized to guide the generator model, and we-energy model is established by the output of the discriminator. This model can accurately demonstrate the static and dynamic characteristic of the terminal integrated energy unit, laying a foundation for the collaborative optimization of the integrated energy system.

Biography

Qiuye Sun (M’11) received the M.S. degree in power electronics and drives and the Ph.D. degree in control theory and control engineering from Northeastern University, Shenyang, China. He is currently a full Professor with Northeastern University and obtained Special Government Allowances from the State Council in China. He has authored or coauthored over 200 papers, authorized over 100 invention patents, and published over 10 books or textbooks. He is an Associate Editor of IEEE TNNLS, IEEE Access, IEEE/CAA Journal of Automatica Sinica, CSEE Journal of Power and Energy Systems, IET Cyber-Physical Systems, Journal of Control and Decision, Mathematical Problems in Engineering. His current research interests include optimization analysis technology of power distribution network, network control of

Energy Internet, Integrated Energy Systems and Microgrids.

DDCLS’20

17


Dynamic Reference Programming-Based Model Predictive Pattern Control by Dynamic Controlled PCA

Prof. Xiaoli Luan Jiangnan University, China


10:20-10:50 VIP23 Hall / VIP23 厅

Abstract A dynamic controlled principal component analysis (DCPCA) algorithm is proposed to extract desirable

latent variables and construct the pattern space of industrial process from a set of measured variables. The constructed pattern space contains the most variations of process variables caused by free motion, as well as the forced movement subjected to the causality originating from control inputs. Consequently, the pattern can characterize the process running state maximally and comprehensively with the minimum dimensions, and the pattern motion equation can be identified to describe the dynamic behavior of industrial plant. After that, a dynamic reference programming-based MPC is designed to drive pattern to track the optimal operation point with zero steady-state error if the target is reachable, otherwise the pattern is steered to a suboptimal but closest position to the target. This MPC strategy is characterized by parameterized reference inputs and enlarged terminal constraint set derived from null space analysis, which could guarantee the maximum reachable optimization area is lossless when solving the objective function.

Biography Xiaoli Luan received the B.Sc. degree in industrial automation from Jiangnan University, China, in 2002; the M.Sc. degree in control theory and control engineering from Jiangnan University, China, in 2006; and the Ph.D. degree in control theory and control engineering from Jiangnan University, China, in 2010. Now she is a professor of the Institute of Automation, Jiangnan University. In 2016, she was a Visiting Professor with the University of Alberta, Canada. Her research interests include robust control and optimization of complex nonlinear systems.

18


Recent Advances in Remaining Useful Life Prediction and Health Management Technology

Prof. Xiaosheng Si Henan Polytechnic University, China


10:50-11:20 VIP23 Hall / VIP23 厅

Abstract Stochastic degradation data analysis is the basic and core component to implement life prognosis and

health management of complex engineering systems. Extensive studies on this subject have been witnessed in the fields of reliability and system engineering. This report will be focused on challenging and fundamental problems in data modeling and model solution for the remaining useful life prediction of stochastic degrading systems. The emphasis will be placed on techniques dealing with linear models, nonlinear model, and switching models. Finally, the future directions will be discussed.

Biography

Xiaosheng Si received the B. Eng., M. Eng., and Ph.D. degrees from the Department of Automation, Rocket Force University of Engineering, Xi’an, China, in 2006, 2009, and 2014, respectively, all in control science and engineering.

He is currently a Professor in control science and engineering with the Rocket Force University of Engineering. He has authored or co-authored more than 50 articles in several journals including European Journal of Operational Research, IEEE Transactions on Industrial Electronics, IEEE Transactions on Reliability, IEEE Transactions on Fuzzy Systems, IEEE Transactions on Systems, Man and Cybernetics—Part A, IEEE Transaction on Automation Science and Engineering, Reliability Engineering and System Safety, and Mechanical Systems and Signal Processing. He is an active reviewer for a number of international journals. His research interests include evidence theory,

expert system, prognostics and health management, reliability estimation, predictive maintenance, and lifetime estimation.

Dr. SI is an Associate Editor of IEEE ACCESS.

DDCLS’20

19


Iterative Learning Control with Incomplete Information Prof. Dong Shen

Renmin University of China, China

Sunday, Nov. 22, 2020 11:20-11:50


Abstract Iterative learning control is an effective control strategy for repetitive systems by utilizing the input and

output information of the previous iterations. It has been shown advantageous in dealing with high nonlinearity and complexity while achieving good tracking performance of high precision. In this talk, we will report recent advances in iterative learning control with incomplete information. Here, incomplete information is generally caused by various practical issues such as data dropout, quantization, varying trial lengths, and communication constraints. The control design and analysis under these issues will be elaborated.

Biography

Dong Shen received the B.S. degree in mathematics from Shandong University, Jinan, China, in 2005. He received the Ph.D. degree in mathematics from the Academy of Mathematics and Systems Science, Chinese Academy of Sciences (CAS), Beijing, China, in 2010.

From 2010 to 2012, he was a Post-Doctoral Fellow with Institute of Automation, CAS. From 2016 to 2017, he was a visiting scholar at National University of Singapore. From 2019 July to August, he was a visiting scholar at RMIT University. From 2012 to 2019, he was with College of Information Science and Technology, Beijing University of Chemical Technology, Beijing, China. Since Dec 2019, he has been a Full Professor with School of Mathematics, Renmin University of China, Beijing, China.

His current research interests include iterative learning control, stochastic control and optimization, machine learning and its applications. He has published more than 110 refereed journal and conference papers. He is (co-)author of four monographs, published by Springer, Wiley, and Science Press, respectively. Dr. Shen received IEEE CSS Beijing Chapter Young Author Prize in 2014. He is a Senior Member of IEEE.

21

2020 IEEE 9th Data Driven Control and Learning Systems Conference

(DDCLS’20)

Technical Program and

Book of Abstracts

DDCLS’20

23

Saturday, 21 November, 2020 SatA01 Room 1 Data driven control 13:30-15:30 Chair: Na Dong Tianjin Univ.CO-Chair: Quan Quan Beihang Univ.

13:30-13:50 SatA01-1 Data Driven Control for a Class of Nonlinear Systems with Stochastic Fading Channels Wei Yu Henan Polytechnic Univ.Xuhui Bu Henan Polytechnic Univ.Yanling Yin Jiaqi Liang

Henan Polytechnic Univ.Henan Polytechnic Univ.

This paper investigates the data driven model free adaptive control (MFAC) problem for a class of non-affine nonlinear systems with stochastic fading channels. Firstly, the phenomenon of signal fading is regarded as an independent stochastic process occurring at the output side, which has known mathematical expectations. Using an innovative linearization method, the considered non-affine system is converted into a linear model with a time-varying parameter called PPD and the MFAC controller is redesigned by utilizing the faded outputs. The stability of the system is analyzed rigorously and the influence of incomplete signal transmission on system convergence is explored. Finally, a numerical example shows the validity of the presented strategies.

13:50-14:10 SatA01-2 Data-Driven Stability Margin for MIMO Systems Jinrui Ren Beihang Univ.Quan Quan Beihang Univ.

The notion of stability margin (SM) plays an important role in control engineering. For multiple-input multiple-output (MIMO) systems, the classic SM is no longer applicable. Although some robust SM analysis methods are popular among multivariable systems, they are model-based, or not easy-to-use in engineering sometimes. In this paper, L2 gain margin and L2 time-delay margin are defined for linear MIMO systems, and a corresponding SM analysis method is proposed by utilizing a loop transformation and the small-gain theorem. Most importantly, a data-driven method for measuring the defined SMs is also presented. As a frequency-domain method, this method can be used to obtain the SMs of MIMO systems experimentally on model-free occasions. The proposed SM analysis and measurement method is simple and practical. Simulation and experiment are given to illustrate the effectiveness and practicability of the proposed method.

14:10-14:30 SatA01-3

A Variable Parameter Model-Free Adaptive Control Algorithm and Its Application in Distillation Tower System Yu Feng Tianjin Univ.Na Dong Tianjin Univ.Yongzhou Li Tianjin Univ.Wenjin Lv Tianjin Univ.

In order to achieve better control performance of chemical process, model free adaptive control (MFAC) scheme is improved by adding two new parameters L1, L2, furthermore to apply in distillation tower system. Compared with basic MFAC, the number of parameters in this novel method is reduced and variable. Firstly, nonlinear system with time-varying desired output is used to carry out numerical simulation for the sake of verifying the effectiveness of this algorithm. After that, the improved MFAC algorithm is applied to the control of the distillation tower system, and the result fully demonstrates the proposed algorithm has strong stability, fast tracking speed. At last, for many systems with time delay in chemical process, such as distillation tower system, a set of validated control method frameworks is proposed in this paper. It is expected to be universally popularized and applied to the control of chemical process.

14:30-14:50 SatA01-4 A Method for Analyzing the State Controllability of Linear Discrete Time-varying Time-delay Systems Zhuo Wang Beihang Univ.

Beijing Academy of Quantum Information Sci.Qi Yuan Beihang Univ.

The state controllability of time-delay systems is important for a wide range of scientific and industrial processes. However, few researches up to now have been carried out for extensive studies on this problem. This paper develops a method for analyzing the state controllability of linear discrete time-varying time-delay systems. By establishing an augmented state-space model of the original system, only a few parameters are needed to complete the determination of the state controllability, which greatly reduces the amount of calculation. Then, a specific example is presented to show the effectiveness of the proposed analysis method.

14:50-15:10 SatA01-5 Model Free Adaptive Pitch Control of a Flapping Wing Micro Aerial Vehicle with Input Saturation Tianhe Wang Beijing Jiaotong Univ.Shangtai Jin Beijing Jiaotong Univ.Zhongsheng Hou Beijing Jiaotong Univ.

Qingdao Univ.

In this paper, the dynamics of a flapping wing micro aerial vehicle is analyzed. Aiming at the difficulty of

24

controller design caused by the nonlinearity, time-varying and strong coupling characteristics of the micro aerial vehicle, a full form dynamic linearization based model free adaptive control scheme (FFDL-MFAC) is presented to realize the pitch control of the controlled vehicle. In addition, a compensator is introduced to overcome the control input saturation caused by the limitation of the actuator. Simulation results are provided to demonstrate the effectiveness of the proposed MFAC.

15:10-15:30 SatA01-6 Virtual Constraint Force Control for Teleoperation System of Live-Power Line Maintenance Xia Liu Xihua Univ.Chengwei Pan Univ. of Electronic Sci. & Tech. of ChinaYong Chen Univ. of Electronic Sci. & Tech. of China

In order to reduce the risk of the human operator in live-power line maintenance while performing the maintenance tasks accurately and efficiently, a virtual constraint force control strategy for robotic teleoperation system is proposed. The virtual constraint force is generated by the virtual spring which is feedback to the operator’s hand by the master. The motion of the operator's hand can be constrained within the desired range near the target point and meanwhile, the human operator can have the sense of touch from the slave. The proposed control strategy is verified by three experiments on live-power line maintenance tasks including clamping porcelain insulator pin, clamping drop-out fuse insurance, and clamping object on overhead wires. The results show that compared to the traditional method, the proposed control strategy can save the task execution time of live-power line maintenance, reduce the position tracking error between the master and the slave and improve the stability of the system.

SatA02 Room 2 Model-free adaptive control 13:30-15:30 Chair: Xiaoli Li Beijing Univ. of Tech.CO-Chair: Jian Feng Northeastern Univ.

13:30-13:50 SatA02-1 Observer-Based Adaptive Multi-dimensional Taylor Network Control for Nonlinear Systems with Time-Delay Lei Chu Qingdao Univ. of Sci. & Tech.Shuhua Zhang Qingdao Univ. of Sci. & Tech.Mingxin Wang Qingdao Univ. of Sci. & Tech.Shanliang Zhu Qingdao Univ. of Sci. & Tech.Yuqun Han Qingdao Univ. of Sci. & Tech.

Key Laboratory of Measurement & Control of Complex Systems of Engineering

In this paper, an observer-based adaptive Multi-dimensional Taylor network (MTN) controller is proposed for strictly feedback nonlinear systems with time-delay and unmeasurable states. MTNs are utilized to approximate the unknown and desired control input

signals directly instead of the unknown nonlinear functions. Moreover, a linear state observer is designed for estimating the unmeasured states. Based on the backstepping technique, a novel adaptive MTN control strategy with simple structure and good real time property is proposed. The designed controller can guarantee all the signals of the closed-loop system are bounded and the tracking error converges to a small neighborhood of the origin. Simulation results are given to demonstrate the effectiveness of the proposed method.

13:50-14:10 SatA02-2 PH Control of Slurry in Wet Flue Gas Desulfurization Based on Model Free Adaptive Control Jian Liu Beijing Univ. of Tech.Xiaoli Li Beijing Univ. of Tech.Yang Li Communication Univ. of China

In limestone-gypsum wet flue gas desulfurization process, the pH change process of slurry in absorption tower has the characteristics of high nonlinearity, large lag and various disturbances. According to the process of pH control in wet flue gas desulfurization, a model free adaptive control algorithm based on compact form dynamic linearization (CFDL-MFAC) is designed. Then the simulation is carried out with Matlab based on hammestein model of slurry pH control system. It is turned out that CFDL-MFAC algorithm can effectively use the input and output data of the pH control process to realize the tracking control of slurry pH and obtain high control accuracy, which verifies the effectiveness of the method. Compared with PID control, CFDL-MFAC controller can not only obtain better control effect, but also effectively suppress external disturbance.

14:10-14:30 SatA02-3 Event-Triggered Consensus Output Tracking Strategy for Multiagent Systems Utilizing Model-Free Adaptive Control Weizhao Song Northeastern Univ.Jian Feng Northeastern Univ.

In this article, a model-free-adaptive-control-based (MFAC-based) event-triggered (ET) consensus output tracking problem for multiagent systems (MASs) is investigated. The dynamic models of agents are unknown, and only a subset of agents can acquire the reference trajectory. The consensus tracking algorithm is designed by the real-time input/output data and pseudo-partial-derivative (PPD), which is an important parameter of MFAC approach. An output observer is built to design the centralized ET mechanism. Then, the boundedness analysis that the tracking error is uniformly ultimately bounded (UUB) is given. Finally, a simulation experiment is provided to verify the feasibility of the ET consensus output tracking strategy for MASs.

DDCLS’20

25

14:30-14:50 SatA02-4 Model-Free Adaptive Control Based on Neural Network Observer for the Chaotic Power Supply System Ao Bai Northeastern Univ.Yanhong Luo Northeastern Univ.Huaguang Zhang Northeastern Univ.

In this paper, we considered a type of chaotic power supply system and presented a neural network adaptive method with Neural Network Observer (NNO). First, the mathematical model of the chaotic power supply system is summarized. Then aiming for the unknown model of n-order nonlinear system, the controller is designed by the neural network adaptive method. There is no need to know the accurate mathematical model and state information of the controlled object. We estimate the state information and model information of the controlled object through the input and output data of the object first, and use the obtained estimation results to implement the controller, and give the corresponding theoretical analysis. Finally, the effectiveness of the designed controller is verified by simulation of a power system with chaotic motion.

14:50-15:10 SatA02-5 Model Free Adaptive Control for the Temperature Adjustment of UGI Coal Gasification Process in Synthetic Ammonia Industry Shida Liu North China Electric Power Univ.Jiao Sun North China Electric Power Univ.Honghai Ji North China Electric Power Univ.Zhongsheng Hou Qingdao Univ.Lingling Fan Beijing Information Sci. & Tech. Univ.

In this manuscript, a data-driven model free adaptive control (MFAC) method is introduced for a UGI gasifier. During the UGI gasification process, the temperature of crude gas inside the UGI gasifier is very important. However, the accurate multi-input and multi-output mathematical model describing the dynamics of the crude gas temperature cannot be created due to the complexity of the gasification systems. The main feature of MFAC method is that the controller design depends only on the input and the output measurement data of the controlled plant. Specifically, the MFAC controller is designed via a novel dynamic linearization technique with a time varying parameter termed Poseudo-Jacobian Matrix (PGM), which contains the coupling information of each output variable. Further, simulation results show that MFAC has a very reliable tracking ability for the temperature adjustment of the gasification process.

15:10-15:30 SatA02-6 Adaptive SMC-based Trajectory Tracking Control of Underactuated Overhead Cranes Shengzeng Zhang Zhejiang Univ. of Tech.

Singapore Institute of Manufacturing

Tech.Xiongxiong He Zhejiang Univ. of Tech.Haiyue Zhu Singapore Institute of Manufacturing

Tech.Yuanjing Feng Zhejiang Univ. of Tech.Qiang Chen Zhejiang Univ. of Tech.Zhengyang Zhu Zhejiang Univ. of Tech.Xiaocong Li Singapore Institute of Manufacturing

Tech.

Overhead cranes, which are typically underactuated, are studied systematically nowadays. While, the model widely used in research is ideal. Thus, the corresponding controllers may react badly under external disturbances, unmodeled dynamics and input constraints. To tackle this issue, this paper develops an adaptive version of anti-sway trajectory tracking controller for overhead cranes. First, as to constrained input, we perform a mapping action from the system input to the hyperbolic tangent function. Then adaptation mechanisms are proposed to adjust the modified inputs and the system uncertainty. Such a controller achieves precise positioning and swing suppression despite input saturation, system uncertainty and external disturbances. The crane system proves to be dissipative with the proposed controller. The experiments accomplished on a laboratory-size bridge crane reveal that the proposed controller asymptotically stabilizes all system states.

SatA03 Room 3 Data-driven fault diagnosis and health maintenance (I)

13:30-15:30 Chair: Jie Ma Beijing Information Sci. & Tech. Univ.CO-Chair: Guo Xie Xi’an Univ. of Tech.

13:30-13:50 SatA03-1 Fault Diagnosis of Rolling Bearing Based on Improved LeNet-5 CNN Siyu Li Xi’an Univ. of Tech.Guo Xie Xi’an Univ. of Tech.Wenjiang Ji Xi’an Univ. of Tech.Xinhong Hei Wenbin Chen1

Xi’an Univ. of Tech.Xi’an Univ. of Tech.

To solve the problem of fault diagnosis of rolling bearing caused by large amount of data and difficulties of processing those data on to bearing set, based on Convolution Neural Network, a new method of data processing is proposed in this paper. With this method, one-dimensional time domain signal can be transformed into two-dimensional images, which is more suitable for Convolutional Neural Network processing. Meanwhile, the traditional machine learning method has the disadvantage of low robustness and low recognition rate with noise interference. Therefore, based on the feature extraction of Convolution Neural Network, in this paper we proposed an improved LeNet-5 Convolution Neural

26

Network model, that is, adding a convolution layer and a pooling layer to the classic LeNet-5 model. The hidden layer features are extracted by using the trainable convolution kernel, while the extracted implicit features are reduced by the pooling layer, the Soft max classifier is used for classification and recognition of rolling bearing faults. In this paper we verified the effectiveness of the improved LeNet-5 model for fault diagnosis of rolling bearing by using the rolling bearing data to train the classic LeNet-5 model and the improved model.

13:50-14:10 SatA03-2 Satellite MicroAnomaly Detection Based on Telemetry Data Chao Sun 63758 Unit of PLAMingzhang E 63758 Unit of PLAYing Du Guangdong Univ. of Petrochemical

Tech.Chuanmin Ruan 63758 Unit of PLA

The military requirements of space security defense and space fast response are increasingly urgent. Accurate and effective micro anomaly detection of on-orbit satellites is an important technical way of satellites life cycle health management. Under this military background, the micro anomaly detection of the key components of the satellite is proposed and carried out. In order to solve the problems of low diagnostic accuracy of the traditional Voherra series model in satellite telemetry signal micro anomaly detection, o-Voherra series anomaly detection model for the feature extraction of telemetry data based on the optimized sequence model is proposed. Firstly, the feature of satellite telemetry data is extracted by using the constructed optimized sequence model. Secondly, phase space reconstruction of telemetry data after preprocessing and feature extraction. Finally, the telemetry data micro anomaly detection are realized by the proposed o-Voherra series model. Through the remote sensing data experiment of the key components of the satellite after desensitization, the proposed model can accurately realize the micro anomaly detection of the key components of the satellite.

14:10-14:30 SatA03-3 Improved PCA-based Fault Isolation using Sparse Group Lasso Wei Chen China Jiliang Univ.Jiusun Zeng China Jiliang Univ.

Jiangxi Univ. of Finance & EconomicsYifan Li Jiangxi Univ. of Finance & EconomicsShihua Luo Jiangxi Univ. of Finance & Economics

In industrial process control, data-driven fault detection and isolation methods have developed rapidly due to the easy availability of large amount of data. In fault isolation, principal component analysis (PCA) based contribution plot is a standard tool. The problem of PCA

based contribution plot is that they are affected by the so called smearing effect. In fact, industrial process variables can be classified into groups according to their correlation or process structure, hence it is straightforward to consider the group-wise fault isolation problem. This paper introduces the sparse group Lasso as a regularization method to improve the fault isolation ability of PCA based contribution plot. The sparse group Lasso term considers both group-wise sparsity and within-group sparsity. Hence more accurate diagnosis can be obtained. In order to solve the optimization problem of sparse group Lasso, an efficient algorithm based on ADMM (Alternating Direction of Method of Multipliers) is proposed. Application study to the Tennessee Eastman (TE) process shows that the proposed method can better isolate faulty variables than competitive methods.

14:30-14:50 SatA03-4 Feature Extraction of Rolling Bearing Faults Based on VMD and FRFT Lei Jiao Beijing Information Sci. & Tech. Univ.Jie Ma Beijing Information Sci. & Tech. Univ.

The fault signal of rolling bearing is non-stationary nonlinear signal, and it is difficult to extract the feature of weak fault under strong background noise. This paper uses a new filtering method-Fractional Fourier Transform (FRFT). Compared with the traditional Fourier transform (FFT), it can make the time-frequency characteristics of unstable fault signals better displayed and suppress cross-interference. In this paper, the method of feature extraction of rolling bearings combined with Variational Mode Decomposition (VMD) and Fractional Fourier Transform (FRFT) is used. First, the original vibration signal is decomposed by VMD to obtain several intrinsic mode component functions (IMF). The component with the largest correlation coefficient is selected as the optimal component for filtering in the fractional order domain. The 1.5-dimensional envelope spectrum of the filtered signal is analyzed. The frequency value corresponding to the maximum amplitude can be obtained. This frequency value is the fault characteristic frequency of the rolling bearing. The simulation results show that the method can effectively extract the fault characteristic information of the rolling bearing.

14:50-15:10 SatA03-5 Bearings Remaining Useful Life Prediction with Combinatorial Feature Extraction Method and Gated Recurrent Unit Network Li Xiao Wuhan Univ. of Sci. & Tech.Zhenxing Liu Wuhan Univ. of Sci. & Tech.Yong Zhang Ying Zheng

Wuhan Univ. of Sci. & Tech.Huazhong Univ. of Sci. & Tech.

Remaining useful life (RUL) prediction is one of the most

DDCLS’20

27

important technologies to implement the health management and predictive maintenance of rotating machinery. To predict precisely the RUL, a three-stage strategy is proposed. Firstly, twenty-four basic characteristics are extracted from vibration signal, which are reconstructed by combining those basic characteristics with complete ensemble empirical mode decomposition with adaptive noise (BC-CEEMDAN), and then the trend curves are extracted to reduce the fluctuation. Next, the most sensitive features are selected by employing a linear combination of monotonicity and correlation criteria. Finally, by input the selected features into the gated recurrent unit (GRU) neural network, we achieve the efficient health indicator with BC-CEEMDAN-GRU. To verify the effectiveness of the proposed approach, experiment on PRONOSTIA bearing datasets is carried out, and the advantage is emphasized by comparison with the six existing methods.

15:10-15:30 SatA03-6 Sensor Correlation Network Based Anomaly Detection for Thermal Systems on Ships Wei Zheng Sci. & Tech. on Thermal Energy &

Power LaboratoryChina State Shipbuilding Corp. Ltd.

Hongkuan Zhou Sci. & Tech. on Thermal Energy & Power Laboratory

China State Shipbuilding Corp. Ltd.Zhiqiang Qiu Sci. & Tech. on Thermal Energy &

Power Laboratory China State Shipbuilding Corp. Ltd.

Zhiwu Ke Sci. & Tech. on Thermal Energy & Power Laboratory

China State Shipbuilding Corp. Ltd.Mo Tao Sci. & Tech. on Thermal Energy &

Power Laboratory China State Shipbuilding Corp. Ltd.

Zhaoxu Chen Sci. & Tech. on Thermal Energy & Power Laboratory

China State Shipbuilding Corp. Ltd.

In this paper, we propose an approach to handle the anomaly detection for the thermal system on ships by the sensor associated network method. A large number of sensors are placed in different positions of the thermal system. These sensors form a topological network which can represent the operation state of the thermal system. There are both linear correlation and nonlinear correlation between the operating parameters reflected by these sensors. The MAS index from MINE is utilized to represent the correlation information between sensors when the thermal system is in dynamic operation condition. Based on the MAS correlation coefficient, the sensor correlation network is constructed to represent the dynamic operation process of thermal system. Using DBSCAN clustering algorithm,

the large topological network is divided into different subnetworks. When the system is running dynamically, the Manhattan distance between the sub networks can reflect the running state of the system. Based on the distance of sub networks, the similarity of sensor networks with continuous changes is calculated and the similarity correlation sequence is formed. Through the matching between the similarity correlation sequence and the historical experience sequence, we can judge whether the system dynamic condition is abnormal. By the simulation experiment data, we verify the effectiveness of the proposed method.

SatA04 Room 4 IS：RNN for computing and its robotic applications

13:30-15:30 Chair: Long Jin Lanzhou Univ.CO-Chair: Shan Liu Zhejiang Univ.

13:30-13:45 SatA04-1 Discrete-time recurrent neural network for solving discrete-form time-variant complex division Zhenggang Pan Yangzhou Univ.Dimitrios K. Gerontitis Aristotle Univ. of ThessalonikiJian Li Xinyang Normal Univ.

In recent years, recurrent neural network (RNN) model has been widely investigated for time-variant problems. In this paper, we focus on discrete-form time-variant complex division solving. Firstly, based on the traditional Euclid division, we present the problem formulation of time-variant complex division. Then, in the continuous-time environment, time-variant complex division is converted a simple time-variant matrix vector equation equivalently; correspondingly, discrete-form time-variant complex division is converted a discrete-form time-variant matrix vector equation. Secondly, we present different discretization formulas and corresponding different discrete-time recurrent neural network (DTRNN) models that have different accuracy for solving the discrete-form time-variant matrix vector equation. Finally, comparative numerical experimental results are conducted to prove the effectiveness of the DTRNN models for solving discrete-form time-variant complex division.

13:45-14:00 SatA04-2 A Long Short Term Memory Network Based on Surface Electromyography for Continuous Estimation of Elbow Joint Angle Yuanyuan Chai Changchun Univ. of Tech.Keping Liu Changchun Univ. of Tech. Zhongbo Sun Changchun Univ. of Tech.

Jilin Univ.Gang Wang Changchun Univ. of Tech.Tian Shi Jilin Univ.

A simple long short term memory (LSTM) network is

28

built to estimate the model which is described the relationship between the elbow joint angle and the surface electromyography (sEMG) signals in this paper. The sEMG time series of biceps and triceps are the inputs of the model, and the elbow joint angle is the output of the model. The sEMG signals while the user is performing flexion and extension movements are collected by Biopac. Elbow joint angle is measured by angle sensor. The results show that for simple flexion and extension movements, the model is able to eatimate the movement intention of the elbow.

14:00-14:15 SatA04-3 Kinematics Analysis of 7-DOF Upper Limb Rehabilitation Robot Based on BP Neural Network Zaixiang Pang Changchun Univ. of Sci. & Tech.

Changchun Univ. of Tech.Tongyu Wang Changchun Univ. of Sci. & Tech.Shuai Liu Changchun Univ. of Tech.Zhanli Wang Changchun Univ. of Tech.Linan Gong Changchun Vocational Institute of Tech.

To solve the inverse kinematics problem of 7-DOF upper limb rehabilitation training robot, propose a new solution method based on BP neural network. Taking a 7-DOF upper limb rehabilitation training robot as the research object, carry out the forward kinematics analysis, establish the BP neural network model for solving the inverse kinematics and improve the neural network. Finally, MATLAB is used to simulate and verify, the simulation results show that the improved BP neural network model can solve the inverse kinematics of 7-DOF upper limb rehabilitation training robot, avoid the complex problem of traditional inverse solution calculation, and the solution process is simple; compared with the standard BP neural network, the learning convergence speed is faster and the solution precision is higher, so it is a feasible 7-DOF inverse kinematics solution method for upper limb rehabilitation training robot.

14:15-14:30 SatA04-4 Continuous Estimation of Human Knee-Joint Angles from SEMG Using Wavelet Neural Network Wanting Li Changchun Univ. of Tech.Keping Liu Changchun Univ. of Tech.Zhongbo Sun Changchun Univ. of Tech.

Jilin Univ.Gang Wang Changchun Univ. of Tech.Feng Li Changchun Univ. of Tech.Xin Zhang Changchun Univ. of Tech.Yanpeng Zhou Changchun Univ. of Tech.

Surface electromyography (sEMG) signals contain a wealth of information associated with human’s movement. In this paper, a wavelet neural network (WNN) model is proposed and implemented to estimate human knee-joint angles from the sEMG signals. With

the processed signals as input, the WNN model is trained to estimate the knee-joint angles in continuous motion. To validate the effectiveness of the WNN model, one able-bodied person sit in a chair and accomplish leg stretching in the experiment, and simultaneously record the sEMG signals from the vastus rectus (VR) and the angles of the knee joint. Then, the estimation results of the WNN model are compared with the RBF neural network and the BP neural network. The experimental results show that the WNN model has the best performance in the knee-joint angles estimation than the other two neural network models. The root mean square (RMS) error of the knee-joint angles is 6.5054◦ and the time is 5.3271 seconds. The proposed method can be applied to rehabilitation robots or assisted exoskeleton.

14:30-14:45 SatA04-5 On Welding Trajectory Centerline Extraction Based on Fuzzy Neural Network Rong Bai Changchun Univ. of Tech. Changchun Decent Opto-Electronic Tech.

Co., Ltd.Shuaishi Liu Changchun Univ. of Tech.Taiting Liu Changchun Univ. of Tech.

Based on the advantages of visual sensing technology with abundant image information and high measurement accuracy, this paper studies the method of extracting the centerline of the welding trajectory. The image was preprocessed to obtain the grayscale image of the weld, the edge of the weld was detected by fuzzy neural network, the contour image of the weld was obtained iteratively by morphological image processing, and the center line of the weld trajectory was extracted by Hessian matrix. It is verified by experiments that the centerline extraction method of weld trajectory studied in this paper can accurately extract the centerlines of welding trajectories of different shapes.

14:45-15:00 SatA04-6 Power-sum Function Activated Recurrent Neural Network Model for Solving Multi-linear Systems with Nonsingular M-tensor Shuqiao Wang Qinghai Normal Univ.Xiujuan Du Qinghai Normal Univ.

Academy of Plateau Sci. & Sustainability

Recurrent neural network (RNN), as a branch of artificial intelligence, shows powerful abilities to solve the complicated computational problems. Due to the similarity between solving equations and controlling dynamic systems, RNN-based approaches can also be analysed from the perspectives of control. Multi-linear systems, on the other hand, are a type of tensor equations with considerable complexity due to the special structure of tensors. In this paper, a power-sum function activated RNN model is proposed to find the solutions of the multi-linear systems with nonsingular

DDCLS’20

29

M-tensors. It is theoretically proved that the proposed RNN model is stable in the sense of Lyapunov stability theory and converges to the theoretical solution. In addition, computer simulations are provided to substantiate the effectiveness and superiority of the proposed RNN model.

15:00-15:15 SatA04-7 Temporal Convolutional Network Based Short-term Load Forecasting Model Kaiming Gu Shanghai Univ.Li Jia Shanghai Univ.

Load forecasting has always been the focus of energy management system research. Recently, with the development of machine learning and artificial intelligence technology, more and more models are applied to load forecasting. In this paper, we design a model based on the temporal convolutional network for short-term load forecasting, which can accurately capture the feature form historical load data. Combine the actual load data collected from a certain region of Shanghai, we compare our model with three traditional models, including ARIMA model, ANN model, and LSTM model. The experiment results show that the model proposed in this paper achieves the best performance and has superior accuracy in short-term load forecasting.

15:15-15:30 SatA04-8 RGBD Object Recognition and Flat Area Analysis Method for Manipulator Grasping Kaijun Wang Zhejiang Univ.Shan Liu Zhejiang Univ.

Grasping guided by visual recognition and positioning is a practical requirement of discrete automation. This paper proposes a general method of using RGBD image recognition, which can recognize object with only one RGB template photograph of the target object, calculate 3D coordinates and pose, and guide the 6 DOF manipulator to grasp object. SIFT(Scale Invariant Feature Transform) is used to extract feature points of template image and real-time scene image and complete matching. Matched feature points are used as seed points to segment target objects in depth image. This method is fast and requires less prior knowledge. In order to optimize the grasping, this paper uses the Shape Index method to locate the flat area on the object which is most suitable for the suction cup. This method can make the grasping system automatically adapt to various objects and overcome the problems of overlapping and partial occlusion.

SatA05 Room5 IS：Data-driven adaptive control for uncertain nonlinear systems 13:30-15:30

Chair: Bing Song East China Univ. of Sci. & Tech.

CO-Chair: Guanbin Gao Kunming Univ. of Sci. & Tech.

13:30-13:45 SatA05-1 An ESN based Modeling for Roll-to-Roll Printing Systems Zhihua Chen Guangzhou Univ.Tao Zhang Huazhong Univ. of Sci. & Tech.Zheng Zhang Guangzhou Univ.

In this paper, a modeling scheme based on Echo State Networks (ESN) is designed and discussed for modeling in Roll-to-Roll (R2R) systems. R2R system involves transport and process of thin, flexible, continuous materials (called webs). An accuracy model is critical to the research of R2R system, such as model-based control and prediction. Existing mechanism modeling methods currently used in R2R systems require complex derivation and do not provide the accuracy performance for changing operating conditions and material properties. The modeling scheme based on ESN utilizes the nonlinear approximation approach where the optimal output connect weights of the network are calculated based on matching of the actual closed-loop R2R printing system. The model established by the proposed method considers the effect of operating conditions and material properties. Experimental data from an industrial printing system is used to corroborate the accuracy of R2R system model can be raised double by the proposed method which compared with mechanism modeling methods.

13: 45-14:00 SatA05-2 Backstepping Sliding Mode Maneuvering Control for a Class of Surface Ships Jie Ma Dalian Maritime Univ.Junsheng Ren Dalian Maritime Univ.Weiwei Bai Guangdong Univ. of Tech.Hongyi Li Guangdong Univ. of Tech.

This paper investigates the maneuvering problem for a class of surface ships in the presence of wave disturbance. The maneuvering problem involves the geometric task and the speed assignment along the path. In order to solve the maneuvering problem, a backstepping sliding mode controller are designed under the wave influence. By establishing the conversing law between the thrust and propeller rotational speed, the rotational speed and rudder angle are taken as the controller output signal. Simulation results verify the controller is valid.

14:00-14:15 SatA05-3 A New Compound Fault Diagnosis Method for Gearbox Based on Convolutional Neural Network Mingxuan Xia Nanjing Univ. of Aeronautics &

30

AstronauticsZehui Mao Nanjing Univ. of Aeronautics &

AstronauticsRui Zhang ZhenDui Industry Artificial Intelligent

Co., Ltd.Bin Jiang Jian Huang Muheng Wei

Nanjing Univ. of Aeronautics & Astronautics

ZhenDui Industry Artificial Intelligent Co., Ltd.

This paper focus on the fault diagnosis problem for the compound faults of rotating machine, in which the rolling bearing and the sun gear faults simultaneously occurred are considered as the compound fault. Considering the traditional compound fault diagnosis methods usually utilize the manual fault features extraction, which are mainly dependent on engineering experience, we propose a compound fault diagnosis method named multi-sensor based convolutional neural network (MCNN). For vibration signals of compound faults, the different transmission paths and the positions of the sensors means one part of the embedded single faults may have higher energy. The vibration signals collected from three sensors at different positions can help guarantee the completeness of the characteristics of the compound fault. Then, the multi-sensor signals are combined together and fused by the convolutional operation of the convolutional neural network (CNN) model. The CNN model, which can automatically extract features from the vibration signals and achieve classification, is used for fault extraction and fault recognition. The experiments are presented on the physical platform of power transmission, and the proposed fault diagnosis method can be verified with the satisfied performance.

14:15-14:30 SatA05-4 Multi-source Heterogeneous Data Fusion Method for Pipe Gallery Condition Monitoring Gang Wang State Grid Hebei Electric Power Co., Ltd.Jingwen Liu State Grid Hebei Electric Power Co., Ltd.

Xiong’an New District Power Supply Co.Guopeng Li State Grid Hebei Electric Power Co., Ltd.

Xiong’an New District Power Supply Co.Zhilei Li State Grid Hebei Electric Power Co., Ltd.

Xiong’an New District Power Supply Co.Zhidan Gong Xiamen Great Power Geo Information

Tech. Co., Ltd.Wenlin Huang Xiamen Great Power Geo Information

Tech. Co., Ltd.Helan Wang Xiamen Great Power Geo Information

Tech. Co., Ltd.Guoyuan Cai Shanghai Guyuan Electric Tech. Co., Ltd.

In view of the exponential growth of the pipeline inspection data volume, the lack of utilization and analysis of the data, this paper proposes a method named subspace principal component analysis (SPCA) for pipe gallery condition monitoring that integrates

multi-source heterogeneous data, aiming to improve the intelligent operation and maintenance level of pipe gallerys. First, interconnected distributed heterogeneous data sources are fused into a unified data set based on the JSON-based middleware method. Second, in order to reduce the complexity of condition monitoring and improve the accuracy, data with similar characteristics are assigned to the same subspace. Then, in each subspace, the principal component analysis (PCA) method is used to mine information and extract features. Furthermore, the features of each subspace are fused, and the local outlier factor (LOF) method that does not require data distribution is used to construct the condition monitoring model and analyze the running state. Finally, the effectiveness and superiority of the proposed method are illustrated by testing it on the operation and maintenance data of the pipe gallery and comparing it with the classical methods.

14: 30-14:45 SatA05-5 IBLF-Based Adaptive Finite-time Neural Backstepping Control of An Autonomous Airship With Full State Constraints Yan Wei Shanghai Jiao Tong Univ.Pingfang Zhou Shanghai Jiao Tong Univ.Yueying Wang Shanghai Univ.Dengping Duan Shanghai Jiao Tong Univ.Weixiang Zhou Shanghai Jiao Tong Univ.

This paper investigates the finite-time attitude tracking control problem of an autonomous airship with uncertainties and full state constraints. An adaptive finite-time neural backstepping control approach is designed by using integral barrier Lyapunov functionals. Radial basis function neural networks are applied to model the uncertainties. A finite-time convergence differentiator is introduced to estimate the time derivative of virtual control law. The stability analysis shows that all the closed-loop signals of airship system are bounded, the state constraints are not violated, and the convergence of attitude tracking error in small neighborhood of the origin in a finite time can be guaranteed. Simulations are performed to verify the effectiveness of the control approach.

14:45-15:00 SatA05-6 Dynamic Modeling and Analysis for 6-DOF Industrial Robots Yingjie Li Kunming Univ. of Sci. & Tech.Jing Na Kunming Univ. of Sci. & Tech.Guanbin Gao Kunming Univ. of Sci. & Tech.

The dynamic model of industrial robots is an important part of controllers, which affects the stability and accuracy of industrial robots. In this paper, a dynamic model for 6-degree-of-freedom (6-DOF) industrial robots is established and analyzed. Firstly, the Modified Danevit-Hartenberg (MDH) method is used to build the kinematic model of the industrial robot, and the

DDCLS’20

31

kinematic parameters of the robot are obtained. The kinematic model is also verified in a simulation environment. Then, with the kinematic mode, the relationship between the force and acceleration of a single link is determined using Newton’s equation and Euler’s equation respectively. The speed and acceleration of each link are calculated by extrapolating methods. According to the method of interpolation, the force and torque equations of the joints of the industrial robot are acquired. The force and torque equations of each link of the industrial robot are obtained by internal iterations, and the dynamic equations of the industrial robot are finally obtained by sorting them out. In order to verify the derived dynamic model, a dynamic simulation environment is constructed. The positions, velocities, accelerations and driving torques of industrial robot’s joints under normal working conditions are obtained by trajectory planning. Analysis of the derived data shows that the motors of each joint can meet the driving torque of the robot under normal working conditions, and the positions, velocities, accelerations of the robot are able to meet the design requirement.

15: 00-15:15 SatA05-7 Compound Disturbance Rejection Control for Nanopositioning Using a Phase-Locking Loop Observer Wei Wei Beijing Tech. & Business Univ.Pengfei Xia Beijing Tech. & Business Univ.Zaiwen Liu Beijing Tech. & Business Univ.

In nano-positioning, accuracy and speed are important issues to guarantee the system performance. Integral resonant control (IRC) is able to improve the bandwidth, and phase-locking loop observer (PLLO) based active disturbance rejection control (ADRC) is capable of achieving better closed-loop accuracy. By combining the advantages of PLLO based ADRC and IRC, a compound control technique is proposed. The compound control can deal with hysteresis and vibration, which are main factors affecting the accuracy and speed of a nano-positioning stage driven by a piezoelectric actuator. An identified model of a nano-positioning stage is utilized, and simulations have been performed. Presented numerical results confirm the proposed compound control technology.

15: 15-15:30 SatA05-8 Event-Driven Distributed Kalman-Consensus Filter with Limited Memory Information Chunxi Yang Kunming Univ. of Sci. & Tech.Jie Zhu Kunming Univ. of Sci. & Tech.Chi Zhai Kunming Univ. of Sci. & Tech.

Consider the problem that the estimation accuracy is inversely proportional to the energy consumption in process monitor with wireless sensor networks (WSNs), the concept of the time-efficiency window is proposed based on the relationship between timeliness and

information efficiency. And then, an improved event-driven mechanism combined with this time-efficiency window (TEW) is also designed. Consequently, a novel event-driven distributed Kalman-consensus filter with time-efficiency window (TEDKF) is presented. By adjusting the length of the time-efficiency window, the effective historical data stored in each sensor is used to informati

FINAL PROGRAM and BOOK OF ABSTRACTSFINAL PROGRAM and BOOK OF ABSTRACTS 2020 IEEE 9th Data Driven Control and Learning Systems Conference (DDCLS’20) Liuzhou, China November 20–22,

Documents