Predictions of the Apple Bruise Volume on the Basis of Impact Energy or Maximum Contact Force Using Adaptive Neuro-Fuzzy Inference System (ANFIS)

Abstract

Abstract Fruit quality drops significantly due to physical impacts and contact forces. Stress on the fruit surface during harvesting, transportation and storage operations causes bruising in its tissue and eventually result in fruit failure. Therefore, prediction of the bruise volume caused by impacts can be very important. In this research, adaptive neuro fuzzy inference system (ANFIS) was used to predict the bruise volume caused by the impacts on apples. The input parameters were the maximum contact force or impact energy; curvature radius at the contact point; temperature; and fruit mass. Its response was the bruise volume. The results show that the ANFIS models operated better in the bruise volume prediction than regression models. Between different available ANFIS models, the model based on the grid partitioning showed the best results with a mean squared error of MSE = 0.00015941, which was less than value showed by the sub-clustering mode. However, its implementation time to reach a fixed error was longer. Eventually, impact energy-based models, in contrast to maximum contact force-based models, were more capable in terms of the apple bruising prediction.