Research On Some Problems About Model-Based Diagnosis Of Hybrid Systems

Model based diagnosis build diagnostic model of physical system by connecting relativecomponents, the traditional diagnostic models are usually described by first order logic.According to logic model and inputs of system, the predicted behaviors which supported to beare deduced by logic theory. If the observations of system are conflict with predict behaviors,the system contains failure. By logical reasoning, the component which fails can bedetermined. The whole diagnosing process bases on the model of system and faults, anddiagnostic result will be deduced.Recently, researchers focus dynamic system which consist discrete event system andhybrid system. We focus on model based diagnosis on hybrid systems which have bothdiscrete property and continuous property. The two properties affect each other. Usually, mostof complex dynamic systems have continuous property; it makes researching diagnosis basedon hybrid system meaningful.Some diagnosis methods of hybrid systems are proposed which focus on diagnosabilityand casual incomplete respectively.To hybrid dynamic system, Hybrid Bond Graphs (HBG) is used to modeling system, atthe same time, automaton which describe hybrid system is used. Using automaton model,actual diagnoses is reason. In a controllable system, the faults can be divided to controllerfaults, controlled object faults, sensors faults and execution faults. The former methodsusually consider the execution and controlled faults, but faults can be found at anycomponents such as controller and interfaces. We proposed a diagnosis method which dealsfaults on controller, and give an actual algorithm to solve the problem.Model based diagnosis in distribute system began at about2005, and exist some classic diagnosis and modeling methods. Some of the methods are applicator in system. Distributesystems have simpler model space and every sub system diagnosis in parallel, the methodreduce time and space complexity. Distribute systems don't have global manager, onceproblem can't be solve by local model, communicate will be carried among other subsystems. Based on basic framework of hybrid system, distribute and parallel technologyapplies to faults diagnosis in hybrid system. A diagnosis method of distribute hybrid system isproposed to reduce time and space complexity and promote effect of diagnoses.Diagnosis is usually based on assumption of completeness. That means every behavior insystem has been built in model. When a model is complete, diagnosis can deduce thedeviation of actual behaviors to get a set of abnormal components. But a complete modelwhich contains all behavior is not easy to build. So how to keep diagnosis in incompletemodel is proposed and researched.A diagnosis method under incomplete model is proposed. A diagnostic system whichbuilds on automaton chains and temporal casual graph is proposed. When the mode of systemis changed, new automaton chain will be produced. Once a fault occurs, the local diagnosiswill be carried in automaton chain. An automaton chain based diagnostic method underincomplete model can solve the problem of global undiagnosable which caused by missingmodel. The local diagnosis result can be given based on automaton chain.A premise of model based diagnosis is the candidate model can be diagnosised. Thepremised property is called diagnosability. Diagnosability guarantee the system can becontrolled by model based diagnosis and diagnosis result correctly. Therefore, the research onmodel based diagnosis concludes diagnosability judging. Even at the stage of modeling, therequirement of diagnosability has been contained. To research method of diagnosabilityjudging under different condition of system is basic and important.Parameter division from normal area is easy to find. But the incipient fault is not so. It isthe small state division but affect in long time. The incipient faults don't change the mode ofsystem, but affect the normal behavior of system. The incipient faults can be checked byprediction. We want to check the diagnosability of incipient fault, and give a diagnosis result.The diagnosability has be proved.