Experiments and qualitative analysis by artificial neural network approach on pool boiling of FC-72 on finned surfaces confined by an unheated horizontal wall

Heat transfer measurements were taken during nucleate pool boiling of FC-72 (a saturated liquid utilized for thermal control of electronic components) from extended finned surfaces placed in a horizontal channel. The influence of channel width on boiling behavior for various extended finned surfaces was experimentally investigated. These finned surfaces have an array of uniformly or non-uniformly spaced straight spines with a square cross-section. The fins had a length of 3 mm or 6 mm. Starting from a uniform configuration, for which the width spacing of the spines was 0.4 mm, non-uniform surfaces were obtained by regularly removing some rows of spines. The effects of channel width (0.5 - 20 mm) and spines distribution were analyzed. The case of an extended unconfined surface was investigated and used as a reference to compare the effect of the confinement. An Artificial Neural Network (ANN) with one hidden layer was trained with the experimental data to predict the heat flux. An optimization algorithm was implemented to calculate the geometry that maximizes the predicted heat flux operating near the maximum temperature limit under different constrictions as fixed fin height or fixed channel width.