Optimized distribution of a large number of power electronics components cooled by conjugate turbulent natural convection

Antoine Baudoin , Didier Saury (1, 2) , Cecilia Boström

1 PPrime [Poitiers] - Institut Pprime [UPR 3346]
2 ISAE-ENSMA - École Nationale Supérieure de Mécanique et d’Aérotechnique [Poitiers]

Natural convection allows for passive cooling which is used in many engineering applications. Placing dissipating components on a common vertical heatsink can be optimized to give the best possible cooling capacity. In this study, a numerical model for three-dimensional conjugated convective and conductive heat transfer was used to evaluate the distribution of up to 36 flush-mounted rectangular heaters. The temperature profiles and the heat fluxes were compared with experimental data for validation. The dissipated power was set as an input parameter and the optimal distribution was selected as the one with the lowest temperature elevation. Two different heuristics-a geometric parameter and an artificial neural network-were proposed and evaluated as alternatives to heavy CFD calculations.

Mots clés: Optimization; discrete heating; conjugate natural convection; artificial neural network; Computational Fluid Dynamic

Domaines: Mécanique [physics]; Thermique [physics.class-ph]; Mécanique des fluides [physics.class-ph]

Antoine Baudoin, Didier Saury, Cecilia Boström. Optimized distribution of a large number of power electronics components cooled by conjugate turbulent natural convection. Applied Thermal Engineering, 2017, 124, pp.975-985. ⟨10.1016/j.applthermaleng.2017.06.058⟩. ⟨hal-02134976⟩