DOI: 10.1615/ICHMT.2008.CHT
ISBN Print: 978-1-56700-253-9
ISSN: 2578-5486
INVERSE HEAT CONDUCTION PARAMETER IDENTIFICATION BY GENETIC ALGORITHM BASED METHOD
ABSTRACT
It is a generally applied technique in the determination of thermal properties that the results of temperature measurements are compared to the theoretical solutions of the heat conduction equation under well defined conditions. There are several methods for solving such types of the so-called inverse problems, but none of them is generally accepted, assuredly working. Genetic algorithms have demonstrated their success in solving ill conditioned, complex extreme value problems. According to the preliminary measurement concept, the forward problem refers to the three layered system: hollow cylinder sample, core and shield. All of them regarded as having temperature dependent properties. The solution of the forward problem is based on the finite difference equations that have been solved by the explicit method and by the Crank-Nicolson method with Gauss-Seidel iteration. In order to find the most accurate evaluation with the least CPU time we generated and tested 6 different calculation versions. The results of the forward heat conduction evaluations were verified by analytical and FEM methods. The verifications have shown very good levels of matching both in the explicit and in the implicit C-N method. The searching algorithm (genetic algorithm) in the present status may be applied in the case of linear temperature dependence of the thermal properties. An example is presented for the inverse solution, which proves that the proposed genetic algorithm can be an effective process in determining temperature dependent thermophysical properties.