Abstract
It is expected that lattice-structured heat sinks, which can be fabricated by metal additive manufacturing (AM) technologies, are applied to thermal management components such as heat sinks. It is important to understand the relationship between lattice structures and heat transfer properties, and optimize lattice structures as heat sinks. In the present study, computational fluid dynamics (CFD) simulations were carried out for evaluating the heat transfer properties of various lattice structures under forced convection. The correlations between various lattice structural features and heat transfer properties or pressure loss were analyzed by the random forest to clarify important lattice structural features for these properties. Four parameters describing the bottleneck in the flow pathway, surface area, and intricate structures were selected. Using the selected structural features, neural network surrogate models were constructed for predicting heat transfer properties and pressure loss of lattice structures. Constructed surrogate models precisely predicted heat transfer properties 106 times faster than the CFD model. Using the surrogate models, the unit cell morphology, number of unit cells, and volume fraction of solids were optimized in terms of the balance between heat transfer property and pressure loss.
