Método de otimização de dissipadores de calor aletados utilizando inteligência computacional

Abstract

With the technological advancement of the latest years, electrical and electronic equipments have become ever faster, incorporating new functions and drastically reducing its size, leading to higher power density and heat flux in its components. To keep the operation temperature of these equipments within acceptable levels, without compromising safety nor realiability, the use of finned heat sinks is essential. Within this context, in this Thesis was realized a theorical numerical optimization analysis on the impact of geometric variables such as the height, width and number of straight fins on the thermal performance of finned heat sinks used in the cooling of electrical and electronic equipment. For this analysis, a setup for a straight fins heat sink was validated with experimental data available in the literature and the Generalized Subset Design (GSD) method of Design of Experiment (DOE) was used in the selection of scenarios to be simulated in order to obtain a data bank with lower computational cost. All these cases were simulated using a commercial Computational Fluid Dynamics (CFD) software, and the results were used in training Machine Learning models, using the open source development language Python, with the purpose of predicting the thermal performance results of finned heat sinks. The models chosen were the decision tree and support vector regressors, multilayer perceptron and neural network developed from the ground up using Keras. A theorical analysis of each variable was done and its impacts in each of the tracked results were presented in graphical form. It was found that the number and width of fins has a middle optimal value, between the set extreme points, showing a non-linear behavior. The prediction Machine Learning models were applied in the detection of behavior patterns of each variable and calculation of the effect such geometric parameter have in the thermal performance of finned heat sinks.