Experimental study and application of an artificial neural network (ANN) model on pulsed spray cooling heat transfer on a vertical surface

Abstract An experimental study is conducted on pulsed spray cooling heat transfer on a vertical surface under controlled nozzle pressure. Nozzle flow rate is found to be directly proportional to the duty cycle when the nozzle pressure is maintained constant and hence fixed at 2 bar throughout the investigation. The temperature fluctuates on the heated surface due to the interaction of the injection/non-injection spray cycles. The temperature fluctuation amplitude decreases with an increase in the distance from the heated surface. However, the temperature fluctuations along three measuring locations from the heated surface disappear when the spray frequency is equal to or larger than 5 Hz. It is found that the heat flux decreases as the spray frequency decreases at a fixed duty cycle. The influence of the spray frequency on the heat flux is more significant at a relatively small duty cycle. The heat flux decreases with a decrease in the duty cycle in both the single-phase and the nucleate boiling regimes. This is mainly due to the decrease in the spray mass flow rate as the duty cycle decreases. The decrease in the heat flux as the duty cycle decreases is more significant in the nucleate boiling regime than that in the single-phase regime. The specific heat flux is used to evaluate the efficiency of pulsed spray cooling heat transfer. It is found that the pulsed spray cooling has a larger specific heat flux than the continuous spray cooling. The pulsed spray cooling is therefore more efficient in fluid usage than continuous spray cooling. Experimental results in this study compare reasonably well with previous correlation in the single-phase regime. However, the comparison is poorer in the nucleate boiling regime. Therefore, a new correlation is developed for pulsed spray cooling on a vertical surface in the nucleate boiling regime. The new correlation predicted experimental data in this study with an accuracy of 10% in the nucleate boiling regime. Finally, an ANN model has been successfully developed for pulsed spray cooling heat transfer on a vertical surface. The ANN model is in excellent comparison with the experimental results. Compared with the empirical correlations, the ANN model shows much higher accuracy in predicting the pulsed spray cooling heat transfer. It is thus concluded that ANN is a promising method which can be integrated in an active spray control system for superior control accuracy.