A priori model reduction method for thermo-mechanical simulations
Résumé
We developed an algorithm to extract reduced-order model from Finite Element model, in the framework of thermo-mechanical simulations. This is an a priori approach in so far as we do not use the state forecasted by the Finite Element model. The basic functions of the reduced-order model are defined for a large evolutions of the state variables of the studied system. An iterative non-incremental algorithm, a LATIN algorithm, constructs them. As the basis functions are improved during the iterative procedure, an approximate state evolution is also bettered. Thanks to the reduced-order model, an accurate overview of the variables evolutions is quickly obtained over the entire time interval. This way we hope to forecast process defaults as quickly as they are significant. An important advantage of this model reduction method is that we do not need a special reflection of what could be the main phenomena that happen during the case studied.