An Artificial Neural Network (ANN) Model to Forecast Onion Shelf Life

Abstract

Prediction of the shelf life of onions using artificial neural network was carried out to obtain high quality product that will last for a long time. Drying of food products is a significant process in food processing and agricultural industry and was dried at varying temperatures (40 – 80 °C).  It is very vital as improper drying conditions can affect the composition of the dried onions and thereby increases the risk of it running into deterioration of the nutrient contents of the product and post-harvest loss. Onions were processed by washing and sorting into required weight. Samples obtained were dried in triplicate using a DHG-9109 oven at different temperatures of 40, 50, 60, 70 and 80oC. A multilayer feed forward neural network was proposed to accurately predict the shelf life of onion. The proposed artificial neural network model was developed to predict the shelf life of onion when supplied temperature, drying time and moisture content as inputs. The oven drying of onions was carried out experimentally at five oven temperatures settings of 40, 50, 60, 70 and 80oC. A number of network configurations consisting of different number of hidden neurons and transfer functions were investigated in this study. Judging from the results of the trained networks, the best performing network was composed of ten hidden neurons and used the tangent sigmoid transfer function. The developed model predicted very accurately and the minimal error, with an overall correlation coefficient of 0.99462 and mean squared error (MSE) of 0.00089. The model also returned regression coefficient (R2) value of 0.9891 for the correlation between the predicted and experimental outputs. The result showed the good generalization of the developed model. These statistical values enhance the reliability of the model for predicting moisture content of onions in food product processing. This work can determine the shelf life of onions and hence calculate its moisture content. In the process of putting together this work, various challenges were encountered in the course of this study, for the aim of future improvements on the ANN for the prediction of shelf life of onion development.