Optimizing Cylindrical Heatsink Design Topology with Feedforward Neural Networks and NSGA II

Authors:

Abstract
Thermal management is a crucial topic in power electronic and will be stressed more and more in the future. This will increase the need for better geometry design of heatsink. This paper discusses the use of Artificial Intelligence (AI) to optimize heatsink designs for power electronics, utilizing a Feedforward Neural Network (FNN) and the Non-dominated Sorting Genetic Algorithm II (NSGA II). It focuses on improving thermal management by accurately predicting heatsink temperatures with input parameters such as power input, pin height, radius, fan speed, and ambient temperature. The findings demonstrate that the neural network’s coefficient of determination (R2) exceeds 0.95, and the Pareto front generated by NSGA II reveals a relationship among the input parameters and heatsink temperature. Furthermore, the most optimal solution from Pareto front was identified by an optimal solution called LINMAP leading to the construction of a heatsink with precise specifications. Another significant contribution of this paper is the discovery, through the analysis of three-dimensional and two-dimensional Pareto fronts, that height has the greatest impact on temperature, followed by the radius of the pin fins, and lastly, fan speed. This finding provides valuable insights for the future design of heatsinks.

This publication has 13 references indexed in Scilit: