Applications of Artificial Neural Network in Determining the Mechanical Properties of Melon Fruits

Abstract

The paper presents the application of artificial neural network (ANN) in predicting some mechanical properties of Melon fruits. There is no established numerical relation between the physical properties and the mechanical properties of melon fruits. The physical data were obtained experimentally for 120 melon fruits. A feedforward backward propagation (ff-BP) architecture was developed and used to estimate the mechanical properties of the Melon fruits from the measured physical properties. A Levenberg-Marquoardt (LM) optimization was used to train the 4-5-5-1 Multi-layer Perceptron (MLP) network and root mean square error (RMSE) was used as performance criteria. The physical properties were trained using MATLAB Neural Network Toolbox and the results compared with experimental values. It was observed that the ANN’s predicted values and the experimental results agreed upto 95%. The network predicts a maximum shear force of 400.76N and a minimum 206.48N. and a best evaluation performance was reached after t epoch three (3).