Identification Des Rdps: Application A La Fiabilite Et Au Diagnostic Paperback

Souleiman Ould El Mehdi

Stochastic Petri nets are powerful tools for modeling industrial processes as well as defects and recoveries that affect them. In this thesis we are interested in the identification of the structure and parameters of stochastic and stochastic-deterministic Petri nets (including exponential and constant crossing times) from the sequences of events dated and recorded by the control systems. supervision. A supervised learning algorithm is developed to build the model structure. This algorithm is based on the back propagation of an error associated with the causality of the events. Then, a method of identifying the parameters is proposed. This method performs a statistical analysis of the content of the sequence and exploits a Markov model isomorphic to the graph of reachability of the Petri net. Finally, built models are used for reliability studies as well as for detecting and locating defects.

9786131557200

French

Omniscriptum

180