An on-line scheme to fault detection in adaptive controlsystems is proposed by introducing Kullbaek Discrimination Information (KDI) as a detection index. When a physical parameter change due to a failure has occurre...
详细信息
An on-line scheme to fault detection in adaptive controlsystems is proposed by introducing Kullbaek Discrimination Information (KDI) as a detection index. When a physical parameter change due to a failure has occurred in a system under adaptive control, the failure effect will hardly be visible in the output performance because of the adaptation mechanism. Such a parameter change is also difficult to detect by monitoring the regulator parameters, which are determined by a recursive identification based on the direct approach. Since the failure effect is reflected as a change in the predictor model used for the adaptive control design, the fault detection leads to a model discrimination problem. It has been shown that the KDI can be used as an effective distortion measure for the model discrimination and the index can be applied in a modified form to detect a fault on-line. Simulation studies on a second order damped oscillatory system have been carried out to demonstrate the efficiency of the method.
作者：
GERSH, JRThe authoris a principal staff engineer at The Johns Hopkins University Applied Physics Laboratory
where he supervises the AAW Operations Section of the Combat Direction Group. Since joining JHU/APL in 1980 he has been involved in the specification development and testing of advanced surface combat direction systems specializing in the application of rule-based control mechanisms to command and control problems. In 1985-86 he chaired the Doctrine Working Group of the Naval Sea Systems Command's Combat Direction System Engineering Committee. Mr. Gersh served in the U.S. Navy from 1968 to 1977 as a sonar technician and as a junior officer (engineering and gunnery) aboard Atlantic Fleet frigates and as a member of the U.S. Naval Academy's Electrical Engineering faculty. He was educated at Harvard University and the Massachusetts Institute of Technology receiving S. B. S. M. and E. E. degrees in electrical engineering from the latter. He holds certificates as a commercial pilot and flight instructor and is a member of the U.S. Naval Institute the IEEE Computer Society and the American Association for Artificial Intelligence.
For the last four years the most advanced surface combat direction system (CDS) of the U.S. Navy has employed a limited knowledge-based control mechanism. Implemented in the Aegis Weapon System's command and decis...
详细信息
For the last four years the most advanced surface combat direction system (CDS) of the U.S. Navy has employed a limited knowledge-based control mechanism. Implemented in the Aegis Weapon System's command and decision element, this capability is called control by doctrine, and is a foundation for the Ticonderoga class cruisers' exceptional performance. control by doctrine allows CIC personnel to direct that certain CDS functions be performed automatically upon tracks with specified characteristics. In effect, these CDS functions, from identification to engagement, can now be controlled through the specification and activation of general system response rules rather than by individual operator actions. The set of active rules, called doctrine statements, forms a system knowledge-base. The Advanced Combat Direction System, Block 1, successor to today's Naval Tactical Data System, will also employ control by doctrine. As part of a larger effort investigating Aegis/ACDS commonality issues, a Doctrine Working Group was chartered to consider, among other things, implications for force-wide interoperability of multiple systems with such rule-based control mechanisms. The working group produced a set of design objectives for doctrine statement standardization across CDSs. Principal features of these objectives are described. The prospect of several such ships operating together in a battle group has raised questions as to the methods by which the actions of ships with those doctrinally-automated systems can best be coordinated. Related questions deal with specific design features for the support of such coordinated action. Work is now being carried out to investigate these questions. Combat system automation through doctrine statements is only one kind of rule-based control. Much work in the area of artificial intelligence deals with the use and maintenance of complex systems of rules, usually in non-real-time problem solving applications. Such systems are just now beginning
A sufficient condition is derived for the existence of an output constant gain feedback controller which stabilizes a single-input single-output dynamical system with a linear nominal part and matched uncertainties. T...
详细信息
A sufficient condition is derived for the existence of an output constant gain feedback controller which stabilizes a single-input single-output dynamical system with a linear nominal part and matched uncertainties. The sufficient condition is significantly less restrictive than a recently derived one [1]. If such a controller exists, it may readily be computed.
The series Advances in Industrial control aims to report and encourage technology transfer in control engineering. The rapid development of control technology has an impact on all areas of the control discipline. New ...
详细信息
ISBN:
(数字)9781846283345
ISBN:
(纸本)9781852339821;9781849969895
The series Advances in Industrial control aims to report and encourage technology transfer in control engineering. The rapid development of control technology has an impact on all areas of the control discipline. New theory, new controllers, actuators, sensors, new industrial processes, computer methods, new applications, new philosophies , new challenges. Much of this development work resides in industrial reports, feasibility study papers and the reports of advanced collaborative projects. The series offers an opportunity for researchers to present an extended exposition of such new work in all aspects of industrial control for wider and rapid dissemination. In some areas of manufacturing, the elements of a flexible manufacturing system form the key components of the process line. These key components are four-fold: a set of programmable robots and machines, an automated materia- handling system that allows parts to be freely routed and re-routed, a buffer storage system where parts and partly-assembled components can wait until required for further processing and assembly and finally, a supervisory control system. The technology employed to coordinate and control all these components as a working system is usually based on programmable logic controllers. The use of this automation hardware and software in manufacturing is designed to yield significant cost reductions and to enhance quality.
Smart grids have become an emerging topic due to net-zero emissions and the rapid development of artificial intelligence (AI) technology focused on achieving targeted energy distribution and maintaining operating rese...
详细信息
Smart grids have become an emerging topic due to net-zero emissions and the rapid development of artificial intelligence (AI) technology focused on achieving targeted energy distribution and maintaining operating reserves. In order to prevent cyber-physical attacks, issues related to the security and privacy of grid systems are receiving much attention from researchers. In this paper, privacy-aware energy grid management systems with anomaly detection networks and distributed learning mechanisms are proposed. The anomaly detection network consists of a server and a client learning network, which collaboratively learn patterns without sharing data, and periodically train and exchange knowledge. We also develop learning mechanisms with federated, distributed, and split learning to improve privacy and use Q-learning for decision-making to facilitate interpretability. To demonstrate the effectiveness and robustness of the proposed schemes, extensive simulations are conducted in different energy grid environments with different target distributions, ORRs, and attack scenarios. The experimental results show that the proposed schemes not only improve management performance but also enhance privacy and security levels. We also compare the management performance and privacy level of the different learning machines and provide usage recommendations.
暂无评论