Study on viscoelastic behavior of pomegranate using artificial neural network and mathematical models

Stress relaxation is one of the defined tests to characterize the viscoelasticproperties of agricultural materials. Stress relaxation data are very importantbecause they provide useful and valuable information such as fruit firmness andripening, food processing and predicting changes in the material during mechanicalloading. Viscoelastic behavior of pomegranate that is cultivated in Iran has beenstudied in current research. For this purpose, stress relaxation test was conductedwith Ardestani cultivar of pomegranate for three sizes (small, medium and large)and three different harvesting times. Two common models (the GeneralizedMaxwell model and the Peleg model) were fitted to the experimental data. ForMaxwell model coefficients, values for RMSE, R2 and MAPE less than 0.003,greater than 0.98 and less than 0.385 were obtained respectively. The value ofMAPE has been acquired for all harvesting times and pomegranate sizes except forpomegranate with medium size in first harvesting time. Results showed thatMaxwell model describes the tension relaxation behavior of pomegranate withlower error than Peleg model in a suitable way. The ability of artificial neuralnetworks technique as an alternative to predict the viscoelastic behavior ofpomegranate was assessed. This assessment showed that artificial neural networksand Maxwell model can predict the tension more reliably and more precisely thanPeleg model. According to analysis of tension data using models coefficients,pomegranate behaves like solid viscoelastic in every harvesting stage. It also showsthat elastic behavior of pomegranate is more in the third stage of harvesting thanfirst and second stages.