Research On Diagnosability Of Discrete-Event Systems

As the development of the technology and deepening of researching on static systems,some researchers paid more attention to dynamic systems. DES can be viewed as a bridgebetween static and dynamic systems, and has the important implications for dynamic systemsresearch. The diagnosis of DES is based on event. If an event occurred, it means that one statechanges to another state. Now we have no solution to salve this problem. It’s difficult to getthe diagnosablity of the global model when the scale is so large, therefore a lot of researchersdivide the global model to local model through rules, and do the diagnosis to the local model.For local model, the typical diagnosis method is distributed diagnosis and decentralizeddiagnosis.This paper focuses on the discrete event system diagnostic, when the system is so large,we cannot solve the problem by global diagnosis easily, we can give a way to divide theglobal model into local models according to some rules, and we can get the diagnosis byforward method. If every local model is not be diagnosed, then we do pruning operation tothe local model, only keeping those useful information for diagnostic, we do limitedprocessing after pruning of the local model and get the limited-local model, finally diagnosethe limited-local model. doing limited to local model is actually draws some useful keyinformation from the synchronization of two local models, avoiding two local modelssynchronize operation, can get the diagnosability as the synchronous operation does, so thatwe can save time and improve the diagnostic efficiency. For the diagnosability of the localmodel can use the algorithm that Lina.YE given. the first in the algorithm is to get the localdiagnoser,and then synchronize itself to get the twin-plant model, from which we can get thediagnosability of the local model. if the model is not diagnosed, we can do abstraction to thetwin-plant model, only keeping communication events, and then synchronize the abstractedmodel with other abstracted model to get the global diagnosability. As the abstracted modelwill be very concise, so we can get the diagnosis without synchronizing to the lastcommunication event. However constructing local model diagnoser and synchronizing itselfwill make the scale expand two times. we present a method to reduce a local model to acritical path model, in this step, we can get the local diagnosability, if it is not diagnosed, we get its’ abstracted model, and we can get the global diagnosability by synchronizing theabstracted model. This method is directly abstract a local model to a abstracted-model thatonly contains communication events and fault events, reducing the size of the model, so thatyou don’t have to build local diagnoser and local twin-plant model, and use the forward searchalgorithm to diagnosis the local model and get the local model that only have critical path.And then synchronize the local-abstracted model, it have two results, one is diagnosable as itcan arrived the final state; the other depends on the result that synchronizing two local modelsthrough the critical path. If there is indistinct states appear, then the global model is notdiagnosed, and give the reason that causes the result, otherwise the system is a diagnosis.