Failure modeling procedures using fault trees and petri nets

Abstract

On designing a fault tolerant system, a unified modeling method is required to model the overall system behaviors, i.e. operating and failure behaviors, consistently. The purpose of this theisis is to unify the operating behavior model and the failure model, that is, a systematic construction of fault trees from Petri nets. A proper manipulation of the Petri net places with the information in the structure of the Petri net can lead to an automatic construction of fault trees because petri nets contain the required information to construct fault trees. Boolean algebra which is equivalent to fault trees is used to express the Minimal Path Sets that can be derived from a Petri net. The limitations of fault trees and Petri nets can be overcome when fault trees are constructed from Petri nets systematically and used with the petrinet simultaneously for designing, developing, and modifying the system. The reliability of the Petri net can be easily verified by the evaluation of the fault tree and the critical places of the Petri net can also be identifien, thus the recovery points of the petri nets can be easily located. The unified method can be useful for the design of fault tolerant systems such as parallel computer, computer network, TDX-10, etc.