Optimal Shape Design in Heat Transfer Based on Body-Fitted Grid Generation

Abstract

This paper deals with an inverse steady-state heat transfer problem. We develop in this work a new numerical methodology to infer the shape a heated body should have for the temperature distribution on part of its boundary to match a prescribed one. This new numerical methodology solves this shape optimization problem using body-fitted grid generation to map the unknown optimal shape onto a fixed computational domain. This mapping enables a simple discretization of the Heat Equation using finite differences and allows us to remesh the physical domain, which varies at each optimization iteration. A novel aspect of this work is the sensitivity analysis, which is expressed explicitly in the fixed computational domain. This allows a very efficient evaluation of the sensitivities. The Conjugate Gradient method is used to minimize the objective function and this work proposes an efficient redistribution method to maintain the quality of the mesh throughout the optimization procedure.