Title: Induction heating processes modelling: optimisation procedure and parallel computing

Authors: Francois Bay, Valerie Labbe, Yann Favennec

Addresses: Centre de Mise en Forme des Materiaux, Ecole des Mines de Paris, B.P.207, 06904 Sophia Antipolis Cedex, France. ' Centre de Mise en Forme des Materiaux, Ecole des Mines de Paris, B.P.207, 06904 Sophia Antipolis Cedex, France. ' Centre de Mise en Forme des Materiaux, Ecole des Mines de Paris, B.P.207, 06904 Sophia Antipolis Cedex, France

Abstract: This paper presents work on the direct modelling and optimisation of induction heating processes. We first introduce the direct model and the ultra-weak coupling strategy developed to deal with these problems. The optimisation problem is then introduced. We wanted to determine the optimal process or control parameters (current intensity, frequency, etc.) in order to get the temperature field or temperature evolution as close as possible to prescribed requirements. The optimisation procedure uses an adjoint method to compute temperature sensitivities with respect to the process parameters and a linear descent algorithm to determine the parameters that will minimise the cost function. The computational costs being quite expensive, we have developed and implemented a parallel computation strategy for the direct model and the optimisation procedure. This strategy is based on an SPMD domain partitioning method. Performance measurements in terms of CPU time have been carried out on industrial applications and are presented.

Keywords: domain partitioning; finite element method; FEM; induction heating; inverse analysis; multi-physics coupling; optimisation; parallel computing; process modelling.

DOI: 10.1504/IJMPT.2007.013130

International Journal of Materials and Product Technology, 2007 Vol.29 No.1/2/3/4, pp.52 - 69

Published online: 09 Apr 2007 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article