Application of an artificial neural network for predicting the texture of whey protein gel induced by high hydrostatic pressure

Abstract

The effects of high hydrostatic pressure (HP), protein concentration, and sugar concentration on the gelation of a whey protein isolate (WPI) were investigated. Differing concentrations of WPI solution in the presence or absence of lactose (0-20%, w/v) were pressurized at 200-1000 MPa and incubated at 30°C for 10 min. The hardness and breaking stress of the HP-induced gels increased with increasing concentration of WPI (12-20%) and pressure. Lactose decreased the hardness and breaking stress of the gel. Furthermore, these results were used to establish an artificial neural network (ANN). A multiple layer feed-forward ANN was also established to predict the physical properties of the gel based on the inputs of pressure, protein concentration, and sugar concentration. A useful prediction was possible, as measured by a low mean square error (MSE < 0.05) and a regression coefficient (R 2 ∈> 0.99) between true and predicted data in all cases. © 2013 IFIP International Federation for Information Processing.