Vibrational Bruise Prediction of Harvested Kiwifruits under Transportation based on the BP Neural Network

Abstract

Vibrational bruise is one of the most common mechanical damages of fruit under transportation. Transportation vibrational bruise prediction can provide important theoretical basis for reducing the bruise and package design. However, there is no related research about the vibrational bruise prediction of fruits. In this study, a vibrational bruise prediction model based on BP neural network was established to predict the vibrational bruise of harvested kiwifruit. The inputs of the prediction model included the vibrational acceleration, vibrational frequency and time, and this network model was trained using adaptive learning rate method based on momentum gradient descent, the vibrational bruise deformation of kiwifruit could be predicted. The results showed that the neural network model has a good prediction effect of vibrational bruise deformation, and the average relative error of predicting vibrational bruise of kiwifruit is 1.32%, and the average absolute error was 0.01, and R2 is 0. 9683. It can provide a weighable theoretical and data reference for the food storage and transportation.