OPTIMAL DESIGN OF CONSTRUCTAL CONDUCTIVE PATHWAYS USING MACHINE LEARNING ALGORITHMS

Increasing the heat flux of future microchips requires implementing a reliable cooling system to reduce their operating temperatures. Researching better designs for high thermal conductivity pathways embedded into heat-generating components is also essential due to concerns about dimensions and costs. The present study investigates the Constructal design of highly conductive pathways embedded into a heat-generating piece using a numerical code based on finite element method (FEM) and an optimization process based on machine learning algorithms (MLAs). The inserted high thermal conductivity occupies a fixed volume fraction. Geometrical features of the highly conductive are considered the optimization variables, and minimization of the peak temperature is considered the optimization objective. To accomplish the optimization process, machine-learning algorithms are used and critically compared to determine the most efficient option among the considered ones. Finally, the optimal Constructal designs predicted by the machine-learning approach are compared with the optimal configurations generated by conventional methods.