Diagnóstico de falhas baseado em autômatos temporizados : aplicação em um sistema modular de manufatura

Abstract

The problem of fault diagnosis has been widely discussed by the academic community using the theory of Discrete Event Systems. However, the application of this theory to real systems is a field where there is a lot to be explored. The problem proposed in this work is to diagnose permanent or intermittent failures in devices (sensors and actuators) belonging to stations of a didactic flexible manufacturing system, called Modular Production System - MPS, produced by Festo company. The objective will be achieved through a modeling and simulation that allow for future implementation in the system. Three methods will be presented on fault diagnosis written in UPPAAL software language which is based on the timed safety automata formalism, as proposed by ALUR and DILL (1994) and HENZINGER et al (1994). The first method is an implementation of TRIPAKIS (2002) diagnoser. The other two methods developed in this research are inspired by TRIPAKIS (2002) and are diagnosable by definitions presented in TRIPAKIS (2002) and I-diagnosability presented in Sampath et al. (1995). The strategies for fault detection include the use of a network of timed safety automata, composed of the automaton that describes the process behavior and the diagnosers automata for each type of failure. The diagnosers detect failures from the observation of delays of certain transitions in the automaton G (process) and isolate them through observations of the sensors states. Fault indicators events serve to announce failures and synchronize the automaton G with the diagnosers.