nondestructive method

Pear damage is one of the main causes of the loss of fruit quality. Bruises occur during dynamic and quasi-static loading, which causes damage to the healthy tissue of the fruit. In this research, pears were placed under quasi-static loading (thin edge and wide edge) and dynamic loading. Then they were stored in 5, 10 and 15 days and after each storage period, using the CT-Scan non-destructive technique the bruise percentage was estimated. In this study, multi-layer perceptron artificial neural network (MLP) by 2 hidden layers and 3, 5, 7 and 9 neurons hidden layers was selected for modeling of loading force and storage period to predict bruise rate. The highest R2 values for training and testing for quasi-static loading of thin edge and wide edge in a 9-neural network were training Thin-edge=0.91, test Thin-edge =0.99 and training Wide-edge=0.95, test Wide-edge =0.99. For the dynamic loading of a network with 3 neurons in the hidden layer has the highest value (training Wide-edge=0.98, test Wide-edge =0.99). For learning (9 neurons) quasi-static loading thin edge (7 neurons) quasi-static loading wide edge and dynamic loading (7 neurons) have been the best network. According to the results obtained for R2, RMSE and learning cycle, it can be said that the neural network has the ability to predict the bruise percentage to an acceptable level for pears.