Intelligent modeling of thermohydraulic behavior in solar air heaters with artificial neural networks

The artificial neural network was applied to map the thermohydraulic performance criteria of the solar air heater with constant values of geometric parameters. Along with intelligent prediction of friction factor and Nusselt number, the main idea behind the research was modeling the performance of solar air heater through the temperature and velocity features instead of geometric parameters which cause opportunity for comparison and investigation of the roughnesses based on the Nusselt number and the friction factor and consequently the thermohydraulic performance. Two approaches were selected for ANN methodology: separate networks for each smooth and roughened plate and integrated model for both. Comparison between the predicted and experimental data in all designed networks indicated that the ANN predicts the Nusselt number better than the friction factor. Parameters of the roughened plate were inserted as the test data to the best-determined smooth plate networks in order to generalize the networks. The better result was observed for the Nusselt number of the roughened plate rather than the friction factor, in the procedure of generalizing the smooth plate networks. The least MSE for the Nusselt number and the friction factor of roughened duct in method 1 was 2.4999e−06 and 2.70199e−04, respectively. 1.7681e−06 and 4.0819e−04 were the least magnitudes of the MSE for the Nusselt number and the friction factor related to the smooth plate. The proposed model can be strongly recommended as a replacement for exhaustive and massive experimental procedure in the optimization of solar air heaters.