Development of Optimal ANN Model to Estimate the Thermal Performance of Roughened Solar Air Heater Using Two different Learning Algorithms

In the present study, artificial neural network (ANN) model has been developed with two different training algorithms to predict the thermal efficiency of wire rib roughened solar air heater. Total 50 sets of data have been taken from experiments with three different types of absorber plate. The experimental data and calculated values of collector efficiency were used to develop ANN model. Scaled conjugate gradient (SCG) and Levenberg–Marquardt (LM) learning algorithms were used. It has been found that TRAINLM with 6 neurons and TRAINSCG with 7 neurons is optimal model on the basis of statistical error analysis. The performance of both the models have been compared with actual data and found that TRAINLM performs better than TRAINSCG. The value of coefficient of determination \((\hbox {R}^{2})\) for LM-6 is 0.99882 which gives the satisfactory performance. Learning algorithm with LM based proposed MLP ANN model seems more reliable for predicting performance of solar air heater.