Application of Backpropagation-Artificial Neural Network in Quality Prediction of Irradiated Black Pepper Beef

Abstract

To investigate the effects of different irradiation treatments on the quality of black pepper beef during storage, a backpropagation-artificial neural network (BP-ANN) model for predicting various quality attributes of black pepper beef was developed based on physicochemical indicators. Irradiation at a dose of 3–4 kGy effectively delayed the loss of juice, lipid oxidation, and protein degradation in black pepper beef during storage, maintained its hardness and microstructure, and increased the contents of umami (Asp) and sweet (Gly, Ala and Ser) amino acids. The BP-ANN model was optimized with the juice loss, thiobarbituric acid reactive substances (TBARs) value, total volatile basic nitrogen (TVB-N) content, tropomyosin band intensity ratio, myosin heavy chain band intensity ratio, and total free amino acid content of irradiated black pepper beef as input variables. The ReLU function was used as the activation function, with 14 neurons in the hidden layer and 100 iterations. The results showed that the 6-14-6 BP-ANN model could predict the quality changes of irradiated black pepper beef well, and have great potential in predicting various qualities of irradiated meat products.

Keywords

• black pepper beef; 60co-γ radiation; quality; backpropagation-artificial neural network; predictive model