Confusion Diagnosis And Supervisory Control For Discrete Event Systems By Using Petri Nets

Discrete event systems (DESs) are event-driven dynamical systems whose evolution of states are due to the performance of discrete events and their behavior is therefore characterized by asynchronous or concurrent evolution processes of discrete events. The evolution processes are not determined by differential and difference equations that have been employed in systems and control engineering to model and study the time-driven behavior of a system, which are correspondingly governed by complex operating rules. The fact makes that the planning, scheduling and control of the behavior of a DES become extremely difficult. Fault diagnosis and control of DESs are a research area that is stimulated by the hope of discovering general principles of the assurances for sys-tem security and reliability, which are considered in a wide rang of application domains such as computer and communication networks, automated manufacturing systems, business management systems, air traffic control systems, highly integrated command, control, communication, and information (C3I) systems, advanced monitoring and con-trol systems in automobiles or intelligent buildings, intelligent transportation systems, distributed software systems, and so forth.Conflicts and concurrency are common phenomena in DESs, where conflicts are constructed to deal with the competition of limited resources and concurrent opera-tions can lead to high operational performance by a multi-threaded pattern. From the viewpoints of system logic and system functions, conflicts are the positions of the decision-making in a DES, which can be thought as the system interfaces between a DES and its external environment. Therefore, the occurrence of conflicts in a DES should be reliable and determinate. However, dependent conflicts and concurrency may lead to the incomplete and indeterminate conflicts that produce difficulties for the resolution of conflicts. The faults are called confusions that usually appear in DESs.This thesis addresses the analytical methods, fault diagnosis and supervisory control strategies of confusions based on Petri nets. The main results of this research are as follows:(1) This thesis presents a formalized definition of confusions in DESs modeled with Petri nets. The formalization begins with the discussion of the mathematical expressions and behavior of conflicts and concurrency in Petri nets. The concepts of concurrent degrees and λ-max-concurrency are proposed to limit the size of subnets in a Petri net. Then, two kinds of confusions, usually called conflict-increasing and conflict-decreasing confusions, are formalized by defining them as a special class of marked subnets with respect to conflicting and concurrent limitations of a Petri net. Confusion structures and behavior can be accurately described and analyzed by the proposed confusion definition.(2) Confusion diagnosis and detection principles in DESs are proposed. First, two kinds of subnets NⅠ/-type and NⅡ-type are defined. A necessary and sufficient condition of deciding whether a NⅠ-type (resp. NⅡ-type) subnet is a conflict-increasing (resp. conflict-decreasing) confusion is provided. According to the condition, this thesis de-velops a confusion on-line detection policy that can capture confusions along with the dynamic evolution of a Petri net system. Second, a confusion off-line detection method based on graph theory and the P-invariants of a Petri net is proposed. Confusion struc-tures are captured by mapping a Petri net structure into a relation graph of conflicts and confusions (RGC2). Moreover, a strategy pertaining to P-invariants of a Petri net is developed to decide whether there exists a marking such that a confusion occurs in the Petri net. Off-line policy avoids to compute the reachability graph of a Petri net.(3) Synchronized Petri nets are the extension of Petri nets in which each transition is controlled by an external event. This thesis develops a class of local synchronized Petri net that is called confusion control event reconfigurable (CCR) nets to obtain a confusion on-line control policy. On the other hand, confusion avoidance policies are also developed by using supervisory control, which can off-line generate confusion-avoidance supervisors such that the controlled net of a plant Petri net is immune to confusions.(4) A workflow net (WF-net) of business procedures is usually said to be correct if the WF-net holds for the properties ’soundness’ and ’well-structured’. However, the ex-istence of confusions in WF-nets may lead to nondeterministic conflicting processes that are undesirable in business procedures. The properties ’soundness’ and ’well-structured’ in WF-nets cannot avoid confusions. Conversely, confusions can affect the properties of a WF-net under some certain conditions. Hence, this thesis analyzes confusions in WF-nets and considers the impacts of confusions on the properties of WF-nets.