Aero-Thermal Optimization of a Heat Sink using Variable Neighbourhood Search

Abstract

This paper presents a geometrical optimization of a heat sink modelled using three-dimensional CFD. The heat sink studied is circular with radial inlets and parallel fins. The parameters of the optimization are the different spacings between the fins. The optimization process is multi-objective and uses an aggregated objective function of both the thermal resistance and the pressure drop of the system. To perform the optimization, a relatively new technique has been used called Variable Neighbourhood Search (VNS). The optimization results give several interesting new geometries. In addition, the performances of VNS are measured with two criteria: the speed of convergence and the repeatability between two optimization runs. These performances are good compared to more traditional optimization techniques like Genetic Algorithms.