Diagnosis, Prognosis, and Health Monitoring of Cooperative Multi-Agent Systems (MAS) and Complex Industrial Systems
Dr K Khorasani
Erskine Visiting Fellow, Professor and Concordia University Tier 1 Research Chair, Department of Electrical and Computer Engineering, Concordia Institute of Aerospace and Design for Innovation Concordia University, Montreal, Canada
Time & Place
Fri, 04 May 2018 14:00:00 NZST in Link 309 Lecture Theatre
There has been a growing interest towards development of networked unmanned autonomous systems that can operate without an extensive involvement of humans. The motivations for this interest may be traced to emergence of applications where direct human intervention may not be feasible due to either environmental hazards, extraordinary complexity of the tasks, or other restrictions. These networks may be potentially made up of a large number of dynamical systems (multi-agents), such as Unmanned Aerial Vehicles (UAVs), Unmanned Ground Vehicles (UGVs), or Unmanned Underwater Vehicles (UUVs). Any of these systems can commonly consist of a number of sensors, actuators and decision makers, and therefore the network of these systems is a network of large number of sensors and actuators or as is known in the literature, a system of systems (SoS).
The increasing complexity of industrial systems such as aerospace systems, transportation systems, to name a few, and the cost reduction measures that have affected the manufacturers and maintenance operators are increasingly driving the need for more intelligence and autonomous capabilities and functionalities for diagnosis, prognosis, and health management (DPHM) of these systems. Maintenance cost accounts for a large part of the ownership cost and the current maintenance strategy for most industrial systems is preventive in which maintenance actions are managed along schedules suggested by manufactures. These schedules are based on historical data, empirical knowledge, and tests performed in design processes and have little to do with the actual condition of the system. To reduce the maintenance cost, predictive and condition-based maintenance is desirable in which maintenance actions are performed whenever they are actually needed.
In this talk, we provide an overview on the control of multi-agent networks, their applications, as well as the solutions that are proposed for them. Specifically, this talk provides an overview on recent results and progresses made on multi-agent consensus, by focusing on cooperative control and consensus (formation) control in presence of communication, control implementation, and saturation constraints. We also present results on the active areas of consensus control, network security, and attacks on the agents as well as networks and resilient consensus control strategies that are proposed to handle these challenges. Moreover, we provide a summary of the research outcomes and accomplishments that we have recently achieved and developed in the DPHM domain of multi-agent systems (MAS). The presented results are categorized into three main groups, namely i) model-based approaches, ii) data-driven and computational intelligence-based methods and iii) hybrid methodology, where a hybrid method refers to a scheme that invokes both model- based and data-driven/computational intelligence-based approaches.
Our main objectives have been to develop modularized concepts for autonomous and intelligent health monitoring, diagnosis, and prognosis of MAS to: 1) provide intelligent automated data analysis functionalities and capabilities, 2) reduce maintenance costs and minimize the chances of catastrophic failures through early detection and monitoring solutions, 3) provide a significant reduction in service engineering and maintenance operations that are labor intensive and involve error prone data analysis tasks,
4) address difficult to process problems that cannot be solved quickly or accurately with conventional
diagnosis methods, and 5) to provide a robust, reliable, and accurate monitoring, diagnosis, and validation system that can operate with actual data.
K. Khorasani received the B.S., M.S., and Ph.D. degrees in Electrical and Computer Engineering from the University of Illinois at Urbana-Champaign in 1981, 1982 and 1985, respectively. From 1985 to 1988 he was an Assistant Professor at the University of Michigan at Dearborn and since 1988, he has been at Concordia University, Montreal, Canada, where he is currently a Professor and University Tier I Research Chair in the Department of Electrical and Computer Engineering and Concordia Institute for Aerospace Design and Innovation (CIADI). His main areas of research are in nonlinear and adaptive control, intelligent and autonomous control of networked unmanned systems, fault diagnosis, isolation and recovery (FDIR), diagnosis, prognosis, and health management (DPHM), satellites, unmanned vehicles, and neural network applications to pattern recognition, robotics and control, adaptive structure neural networks. He has authored/co-authored over 450 publications in these areas. He is currently serving as an Associate Editor of the IEEE Transactions on Aerospace and Electronic Systems.