Prediction of exergetic efficiency of arc shaped wire roughened solar air heater using ANN model

Abstract

In present work, Artificial Neural Network (ANN) model has been structured to predict the exergetic efficiency of roughened solar air heater. Arc shaped wire rib roughened absorber plate with relative roughness pitch (P/e) =10, relative roughness height (e/D) =0.0395 and angle of attack (α) = 60o is used in the experiments. Experiments were conducted at Jamshedpur (India) under the local weather conditions for 7 days between 09:00 and 16:00 h in the month of February. Total 105 data samples have been collected from the experiments. By the use of experimental data and calculated values of exergy efficiency were used to develop ANN model. The MLP model has been developed with eight parameters in input layer and one parameter in output layer. Levenberg-Marquardt (LM) learning algorithm has been used to training an ANN model. For optimal topology, 10-16 neurons were used to train the network and found that 14 numbers of neurons with single hidden layer is optimal on the basis of statistical error analysis. The values of R2 for predicted exergetic efficiency is 0.99583 which shows that predicted values are very close to experimental data. The statistical analysis also show that the value of RMSE and COV were 0.021747 and 1.63766 respectively, which are very low as required for accurate prediction. The statistical results show that the proposed MLP model successfully predicts the exergetic performance of roughened solar air heater.