Design of experiments in thermal architecture

Abstract

Thermal architecture choice has a large impact on a products ultimate performance in thermally limited products. Thermal parameters interact, hence Design Of Experiments (DOE), which can include interaction effects, is the preferred explorative strategy. However, DOE has the drawback of potentially resulting in impractically large experimental arrays and correspondingly overly large datasets and large claim on resources. This work describes a multi-step approach to thermal architecture combining Computational Fluid Dynamics (CFD) enabled DOE based exploration and optimization on a reduced set of parameters. A base case is used to identify the critical outputs and a screening DOE is used to identify the most important, vital few inputs. The subsequent optimization DOE is limited in size and the use of multiple small DOEs result in less runs compared to one large DOE. DOE arrays are generated and analysed in a commercial spreadsheet application. The method is demonstrated on an automotive electronics box, a Body Control Module.