Copyright © IFAC System Structure and Control, Bucharest, Romania, 1997 CONTROL OF A VULCANIZATION PROCESS: A GENETIC IMPLEMENTATION. L.Autrique*,1 J.E.Souza De Cursi** * IMP/CNRS, University of Perpignan, FRANCE, 52 Avenue de Villeneuve, 66860 Perpignan Cedex, Te/ : (+33) 4.68.66.22.39, FAX: (+33) 4.68.67.21.66, e-mail: autrique@univ-perp./r ** Institut de Mecanique de Rouen , I.N. S.A de Rouen, Place Emile Blondel BP 8, 76131 Mont Saint Aignan cedex, FRANCE, Tel: (+33) 2.32.95.97.12, FAX: (+33) 2.32. 95 .97.10, e-mail: Eduardo.Sou za@i nsa-rouen.lr Abstract: Control of industrial processes modelized by distributed parameter sys- tems is rather difficult to implement. The increasing development of non deterministic strategies (simulated annealing, stochastic perturbations, genetic strategies, ... ) give a new alternative to classical optimization methods. In this paper, a genetic approach is proposed in order to determine the solution of optimal control problems. The control of a vulcanization process is investigated and an application to environmental sciences is briefly presented and leads to the determination of a spray control. Copyright © 1998 IFAC Keywords: Optimal Control - Distributed Parameter Systems - Genetic Algorithms - Finite Elements 1. INTRODUCTION Process optimization is essential in order to en- sure industries competitivity. For elastomers and composite materials, empirical methodologies do not satisfy performances and qualities required for high-technology developments (see (Hou et al. 1990)). The example or rubber elaboration is quite significant in so far as thermal properties (over- heating, ... ) and mechanical properties (attrition, impact resistance, tensile strength, ...) are closely connected to the reticulation rate evolution which depends on the history of the temperature be- haviour (see (Khouider and Vergnaud 1988)) . 1 This work was partially carried out at "Laboratoire d'Automatique de Nantes - URA CNRS 823" 483 The problem under interest in this paper is the de- termination of an optimal control giving a desired vulcanization degree. The phenomenom complex- ity leads to a system with two state equations linked by a non linear term. The optimization problem is treated as a func- tionnal minimization one. The development of the coupling of purely deterministic and stochastic approaches has improved the interests of classi- cal descent methods. Recently, the emergence of genetic methodologies in the control fields has offered a new alternative for non convex global optimization problems. In the following, results obtained for the deter- mination of an optimal control by a genetic al- gorithm are presented. This paper is a further contribution to (Autrique and Souza De Cursi 1995) devoted to the coupling between purely deterministic and stochastic methods . First of all, a model for the vulcanization process is pro- posed . Secondly, the optimization problem is in-