Inverse identification of thermal parameters using reduced-basis method

Abstract

A novel computational method is presented to inversely identify heat convection constants for complex engineering systems. A reduced-basis approach is developed as a forward solver for heat transfer analysis in order to significantly reduce the computational complexity in each forward analysis, which is crucial for the inverse analysis of complex systems. An intergeneration-projection genetic-algorithm (IP-GA) is introduced as the inverse procedure to speed up the process of finding the desired global minimum of the fitness function of error that leads to the parameters to be identified. As an example, identification of the heat convection constants of a typical microelectronic package are performed using the present method and the performance is examined comparing with other inverse identification methods. It is found that the reduced-basis method combined with the IP-GA significantly outperforms the conventional methods. The proposed inverse identification method is efficient enough even for online analysis of inverse problems due to both the use of the reduced-basis approach and the present inverse procedure. © 2004 Elsevier B.V. All rights reserved.