Incomplete Problem Diagnosis Of Incremental Models In Discrete Event Systems

Model-Based Diagnosis is a kind of intelligent reasoning method，it is also atheory method and application direction of artificial intelligence. Its main idea ismodeling the structures and behaviors of the target system, then combining the systemwith the observation results produced during the actual process of operating system.After conducting common reasoning, thus it obtains the conclusion that whether thereis a fault in the actual running track of the target system.Discrete event system is a dynamic method of Model-Based Diagnosis, basing onthe off-line modeling and the on-line observation，the method carries out the runningtracks and diagnosis results of the target system. In Discrete Event Systems,incremental model diagnosis owns the reusability of method and the application ofinformation. Not every diagnosis should start from the initial state，basing on theexisting diagnosis results， the target system conduct the subsequent diagnosis.Although the incremental model diagnosis can improve the efficiency of fault diagnosis,in the process of incremental diagnosis, the incomplete problem may exist in the targetsystem.Model incompleteness refers that during the operation of target system, some of thestates or events cannot explain some behaviors of the target system according to theoriginal definition of the system model. Model completeness refers that before theoperation of target system, the model contains all the possible states and events. Duringthe process of modeling the actual physical system, the reason that the actual physicalsystem and logic model mapping is not complete, leads to the incompleteness of modelgeneration and makes the diagnosis result deviate and the error occur. Therefore, according to the incomplete problem of independent model, this paperproposes the diagnosis algorithm of incomplete model. On the one hand, the TotallyOrdered observation algorithm (TOO Algorithm) is used to solve the ideal state,where system events are all observable and the sequence of observations detected bysensors is complete and orderly. On the other hand, the Partially Ordered ObservationAlgorithm (POO Algorithm) is used to solve the general case, where unobservableevents exist in the model of target system, the sequence of observations detected bysensors keeps orderly, but not complete necessarily.Then, aiming to the incomplete problem of incremental model, this paper proposesthe incomplete diagnosis algorithm of incremental model. On the one hand, theIncremental Totally Ordered observation algorithm (ITOO Algorithm) is used to solvethe ideal state, where system events are all observable and the sequence of observationsdetected by sensors is complete and orderly. On the other hand, the IncrementalPartially Ordered Observation Algorithm (IPOO Algorithm) is used to solve the generalcase, where unobservable events exist in the model of target system, the sequence ofobservations detected by sensors keeps orderly, but not complete necessarily.To above algorithms， this paper gives out the summary of automata modelinstances, the comparison of operation results, the analysis of the experimental data andthe property of the experiments. From the independent automata model to the chainstructure of incremental model. From the events in the target system are observable tothe unobservable events exist in the system.The layout and structure of this paper has undergone the development from simplecondition to complex one, from the special situation to the general one. It provides thedetailed study of incomplete problem for readers step by step.