Optimized Placement of IC Chips for Enhanced Thermal Cooling: A Hybrid ANN-GA Approach in Numerical Heat Transfer

This study employs numerical modeling to investigate the natural convection heat transfer characteristics of nine aluminum integrated circuit (IC) chips mounted on a substrate board. The objective is to determine the optimal arrangement of these IC chips to enhance thermal performance. A dimensionless parameter (λ) plays a crucial role in this optimization, achieved through a hybrid methodology integrating Artificial Neural Networks (ANN) and Genetic Algorithms (GA). Three-dimensional steady-state numerical simulations using ANSYS Icepak are conducted to evaluate the temperature distributions of the IC chips. The study reveals that the shape, size, and positioning of IC chips on the substrate board significantly impact their operational temperatures. Compared to conventional approaches, the proposed hybrid optimization technique demonstrates superior reliability in accurately predicting the optimal IC chip arrangement. A direct correlation is established between parameter variations and the reduction of maximum temperature excess, highlighting the influence of key design parameters. The most favorable simulation results guide the GA, which identifies an optimal chip configuration that ensures heat sources remain within their specified operating limits, as defined in the manufacturer’s data sheets. The temperature variation between the highest and lowest extreme configurations is found to be within 4-8%. Notably, the smallest IC chip (U2) with high heat dissipation reaches the highest temperature in the setup, whereas low-powered chips (U3, U4, and U7) exhibit minimal temperature variations. The study reports strong agreement between numerical predictions and experimental data, with an error margin within 10%, confirming the accuracy of the developed hybrid optimization model.