Thermal Analysis of a Novel Jet Impingement Solar Air Heater Design Using Optimal Training Function in an Artificial Neural Network

The objective of this study is to model and obtain a predictive thermal analysis of a Jet Impingement Solar Air Heater (JISAH) with a novel conical protruded nozzle design using a Multi-Layer Perceptron (MLP) approach in Artificial Neural Network (ANN). The ANN model is constructed using analytical data obtained from Finite Element Method (FEM) based on numerical analysis of the current novel design of the JISAH. The numerical analysis was performed previously using COMSOL Multiphysics Computational Fluid Dynamics (CFD) software. For the training process of the present ANN model, a feed-forward learning strategy is adopted with four separate training functions, viz. TRAINOSS, TRAINSCG, TRAINCGP and TRAINLM. The predicted values are compared with the results of thermal performance obtained from the CFD investigations. The values of Root Mean Square Error (RMSE) and Coefficient of Determination (R2) are utilized for selection of the most optimum training function based on its approximation of thermal performance for the present JISAH design. It is found that TRAINLM training function is found most suitable for predicting optimal results for absorber plate temperature, useful heat gain and thermal efficiency when compared with the previously calculated CFD results.