Multiobjective optimization of component placement on planar printed wiring boards

Abstract

This paper presents a solution methodology for the optimal placement of convectively and conductively air-cooled electronic components on planar printed wiring boards considering thermal and electrical/cost design objectives. The methodology combines the use of a heat transfer solver for the prediction of the temperature distribution among the electronic components and a genetic algorithm for the adaptive search of optimal or near optimal solutions and a multiobjective optimization strategy (Pareto optimization and Multiattribute utility analysis). After proper validation of the elements of the solution methodology (heat transfer solver/genetic algorithm) in isolation, the methodology under consideration is tested using a placement problem (case study) that considers as optimization criteria the minimization of an estimate of the failure rate of the system of components due to thermal overheating (via an Arrhenius relation) and the minimization of the total wiring length (given some interconnectivity requirements). Results corresponding to the case study are presented and discussed for both Pareto optimization and Multiattribute utility analysis.