Architected materials informatics: Construction and application to cellular-structured heat sink optimization

Abstract

Additive manufacturing permits intentionally controlling the geometry of architected materials to achieve desired properties. It is important to design and optimize the geometry of architected materials within a wide search space. This study integrates Voronoi tessellation and informatics to construct architected materials informatics for geometry optimization. Voronoi tessellation can design diverse architected materials by dividing a design space with bisecting planes between arbitrarily arranged seed points and replacing the bisecting plane edges with solid struts. The neural network surrogate model predicts the properties of the architected materials based on the seed point coordinates and the strut radius. The genetic algorithm inversely analyzes the surrogate model to optimize the seed point coordinates and strut radius directly linked to the optimized architected material geometry. This framework rapidly optimized the geometry of cellular-structured heat sinks, optimally balancing pressure loss and heat transfer (Pareto front) under forced convection. The optimized geometry of the Pareto front was validated using computational fluid dynamics and experiments. In addition, the data obtained during the optimization was analyzed to develop a strategy to improve the pressure loss and heat transfer trade-off. This study provides architected materials informatics framework to optimize the geometry of architected materials for diverse applications.