CHARACTERIZATION OF SPACE DEPENDENT THERMAL CONDUCTIVITY WITH A BEM-BASED GENETIC ALGORITHM

An inverse boundary element method is developed to characterize the spatially dependent conductivity k(x) of heterogeneous materials. Surface measurements at exposed boundaries serve as additional input to a genetic algorithm (GA) minimizing a regularized functional measuring the difference between observed and boundary element method (BEM)-predicted surface temperatures under current conductivity estimates. Analytical integration for a given estimate of k(x) provides explicit evaluation of influence coefficients, and series expansions are used to approximate integrands in expressions for these coefficients. Numerical integration is performed just once, a significant speedup in updates of the functional in evolution of the GA.