Multiobjective Optimization of a Pin-Fin Heat Sink Using Evolutionary Algorithms

This paper presents the use of multiobjective evolutionary algorithms for the optimal geometrical design of a pin-fin heat sink. The multiobjective design problem is posed to minimize two conflicting objectives: the junction temperature and the fan pumping power of the heat sink. The design variables are mixed integer/continuous. The encoding/decoding process for this mixed integer/continuous design variables is detailed. The multiobjective optimizers employed to solve the design problem are population-based incremental learning, strength Pareto evolutionary algorithm, particles swarm optimization, and archived multiobjective simulated annealing. The approximate Pareto fronts obtained from using the various optimizers are compared based upon the hypervolume and generational distance indicators. From the results, population-based incremental learning (PBIL) outperforms the others. The new design approach is said to be superior to a classical design approach. It is also illustrated that the proposed multiobjective design process leads to better design compared to the current commercial pin-fin heat sinks.