Utilization of Artificial Neural Networks in the Context of Materials Selection for Thermofluid Design

Abstract

In this article, we use artificial neural networks (ANNs) to approximate the design space of heat transfer problems involving several choices of materials. The approximations provided by ANNs are used with genetic algorithms (GAs) to optimize the systems. Three test cases with multilayer structures are studied: 1) layered porous media heat sink, 2) finned heat sink, and 3) exterior building wall. Important computational time savings are reported compared to optimizations with GAs that rely on direct simulations. Optimal or nearly optimal designs have been identified in each case.

The authors' work was supported by the Natural Science and Engineering Research Council of Canada (NSERC).

Notes

^a Empty cells represent a layer of insulator.

^b Cells with a number represent a layer of PCM (with specified T _m ).