Abstract : This thesis deals with the contribution of online simulation to production activity control. Along the years; manufacturing systems get more and more complicated and require more and more analysis skills of the pilot. Several studies showed that the most precious help that a decisional entity could get to optimise its decisions was a tool enabling to foresee the consequences of these decisions. Indeed, the rise of systems complexity makes this aspect of decision making particularly tricky. The subject of this thesis is to study how discrete-event simulation, after being used as a support for production units design, may become a decision support tool for production activity control of these units.
First section presents the context of this work. After positioning the manufacturing systems production activity control, we present the discrete-event simulation tools and principles. Finally, we present a state of the art about online simulation, noticing it is a relatively poorly explored subject by the manufacturing systems research community. Second section of our work deals with the integration of simulation tools in the control architecture. Our first step is to study the relative position of human and machine in the decision making process. For a long time, classical simulation tools has enabled to foresee the consequences of decisions acting on a term considered long relatively to the system dynamic. When the horizon shortens, the initial state of simulations becomes a problem: because of its growing influence on the results, it is necessary to insure a more and more strict equivalence with the state of the system at the simulations start date. This is why, in the second part of this section, we show on an example the importance of initialisation in an online simulation for decision making.
Third section of this work is about the resolution of this last problem with the insertion of a simulation-made state estimator in the control architecture, constantly communicating with the system control. At the simulation start date, it initialises on the present state of the observer, which is considered as a good approximation of the real present state of the system. Last section presents how we validated what we proposed with a complete application of the proposed concepts and architecture on a flexible manufacturing system, currently running at the Technology University Institute (IUT) of Nantes, France.