Testing Of Diagnosability Of Discrete-event Systems And Approach Of Diagnosis Of Higher-order Discrete-event Systems

No matter it is a simple physical circuit or a large-scale industrial system, it oftenfails. If we can’t handle the failure timely, the consequence is serious. So equipmentsneed to be diagnosed and monitored. It is easy to identify failure on the simplesystem. But it is hard to identify failure on the large-scale system by human.Therefore automated diagnosis was proposed in order to ensure systems security andreliability. Now, many diagnosis methods have been used in real-world applications,including aerospace equipment, large power transmission networks,telecommunication net, electronic commerce, testing of large software componentsand so on.The principle of faults diagnosis is based on observations on its behavior, todeduce the failure that might happen when the system is running. The technology ofdiagnosis should be able to localize and identify the failure as accurately as possible.Earlier, diagnosis was used on static systems, and then many researchers have begunto research diagnosis of dynamic systems since the last decades. Discrete-eventsystems (DESs) can represent dynamic systems at a higher level of abstraction.Traditional fault diagnosis methods are based on expert knowledge and specifiedsystems. Thus these methods are too restrictive. Model-based diagnosis is a type ofintelligent diagnostic reasoning technology, overcoming shortcomings of traditionalfault diagnosis methods. It is a hot researching topic in artificial intelligence. Modelbased diagnosis method build system’s model according the inner structure andbehavior of the system, to deduce possible failure thanks to observations on itsbehavior. Diagnosis of DESs has been attracting attention of the scientific communitysince Sampath et al. proposed diagnosis problems in1996. Previous research hadmainly focused on relevant yet different aspects, including the global diagnoserapproach, decentralization approach, distribution approach, hierarchical approach,incrementality approach to diagnosis of DESs. However, these methods areirrespective of the context of DESs. So these methods are only used for simplesystems. The behavior of a higher-order DES (HDES) that is used for the diagnosisof complex DESs is stratified and each sub-DES at different levels has its ownbehavioral model. An improved approach of fault diagnosis of HDES and a formalalgorithm were proposed. And the definition of diagnosability and the algorithmused to test diagnosability were presented. To diagnose HDES, it needed to constructdiagnosis engine of HDES which was also used for testing diagnosability. The stateof diagnosis engine included sets of faults. Thus, the faults associated with final stateare the candidate diagnose of the system.Testing diagnosability of DESs is a very important step before diagnosing DESs.It can ensure correctness and uniqueness of the diagnosis result. Diagnosabilitydetermines whether a system can be diagnosed by deducing the failure from asequence of observations. It means that a failure can be detected by using theinformation of the observed events. Now, these methods used for testingdiagnosability need to construct diagnoser or twin plant. But the complexity is toohigh. An improved approach used to test diagnosability was presented. Abstractingglobal model based on the type of states, the result of global diagnosability can begiven according to local diagnosability.