Simulation avancée des problèmes thermiques rencontrés lors de la mise en forme des composites

The modeling of composites manufacturing processes remains today a scientific challen-ging issue despite the impressive progress reached in mechanical modeling, numerical analysis, discretization techniques and computer science during the last decade. Indeed, composite manufacturing involves highly non-linear anisotropic behaviors and strongly coupled multiphysics defined in complex geometries. Moreover, optimization, inverse analysis and process control require the solutions of many direct problems, as fast and accurate as possible. In this context, reduced order models constitute an appealing simulation choice, acce-lerating the computations of several orders of magnitude, and even enabling the solution of models never solved until now. The “Proper Generalized Decomposition” or PGD is one of the three main families of reduced order model techniques. PGD represents a new paradigm in computational mechanics. PGD can address the solution of multidimensional problems involving space, time and parameters as extra-coordinates, while circumventing the curse of dimensionality thanks to the separated representations that it involves. In this work we use the PGD to solve thermal problems encountered in composite forming processes. Moreover, an “offline/online” computational technique is proposed in order to optimize and control processes in real time. In fact, the PGD is used to compute parametric solutions “offline”, while optimization techniques are performed “online” in order to identify optimal material, process or geometrical parameters. Furthermore, “Online” calculations can be performed on light computing devices like smartphones or tablets.