Modelling of a locally fabricated flat-plate solar milk pasteuriser using artificial neural network

Abstract

The objective of this work was to develop an artificial neural network model to predict milk temperature of a locally fabricated solar milk pasteuriser, based on measures of error deviation from experimental data. A three-layer feed-forward neural network model based on back propagation algorithm was developed using the Neural Network Toolbox for MATLAB®. The inputs of the model were ambient air temperature, solar radiation, wind speed, temperature of hot water, and water flow rate through the collector, whereas the output was temperature of milk being pasteurised. The optimal neural network model had a 4-4-1 structure with sigmoid transfer function. The neural network predictions agreed well with experimental values with mean squared error, mean relative error and correlation coefficient of determination (R2) of 5.22°C, 3.71% and 0.89, respectively. These results indicate that artificial neural network can successfully be used for the prediction of the performance of a locally fabricated solar milk pasteuriser.