Shelf Life Prediction of UHT Milk Packaging Based on BP Neural Network

To investigate the effects of initial protein, fat content, and storage temperature on the shelf life of UHT pure milk packaging, three types of UHT pure milk were used as research objects to experimentally measure sample browning index and protein hydrolysis index during storage at 23, 30, and 37 ℃. Integrate the dataset and determine specific input parameters based on its performance on the prediction set, and carry out UHT pure milk packaging shelf life prediction based on BP neural network. The results showed that the fitting degrees of the BP neural network model for the browning index and protein hydrolysis index of UHT milk were 0.9412 and 0.9527, respectively, and compared with traditional multiple linear regression model’s number of 0.8799 and 0.9211, the BP neural network model with optimized hidden layer neuron numbers had higher prediction accuracy for the changes in characteristic indicators during the storage period of UHT pure milk, providing technical support for rapid and accurate prediction of the shelf life of UHT pure milk with different formulas.