Starch-based aerogel prepared by freeze-drying: establishing a BP neural network prediction model

Preparation of starch-based aerogel by freeze-drying technique is considered as a suitable method for practical production due to its reasonable cost and controllable process. However, the low expansion rate of this method limits the application of adsorption and transport. In this study, through some pretreatment methods, such as adding glycerol and NaHCO 3 , and using double-roll extrusion, etc., the volume of the internal pores of starch aerogel was effectively increased, and the expansion rate was significantly improved. Meanwhile, a two-layer BP (back propagation) neural network model was established to predict the expansion rate. The result shows that the established BP network prediction model can accurately predict the target. Meanwhile, as the mass of glycerol increases, the expansion rate first increases and then decreases. When the mass of glycerol reaches 30 g, the expansion rate reaches the maximum value of 13.7%. As the mass of NaHCO 3 increases, the expansion rate also increases first and then decreases. With the continuous increase of the double-roll extrusion time, the expansion rate will continue to increase. The simple pretreatment methods have a significant effect on improving the expansion rate of starch aerogel, which expands the application range. At the same time, the successfully applied BP network prediction model can improve production efficiency. Graphical abstract