BP neural network model for predicting the mechanical performance of a foamed wood‐fiber reinforced thermoplastic starch composite

A foamed starch‐based biomass composite was prepared by the extrusion foaming method with starch, wood fiber and polyvinyl alcohol (PVA) as raw materials. The effects of three additives (plasticizer, cross‐linking agent, and blowing agent) on the mechanical performance of the foamed starch‐based biomass composite were studied, and a three‐layer back‐propagation (BP) artificial neural network model was established for mechanical property prediction. The orthogonal experiment results were used as samples to train the network and predict the mechanical performance of foamed composite. The results show that the proposed BP network model is effective in forecasting the mechanical performance of foamed composite. The appropriate amount of composite plasticizer (glycerin/NaOH) can reduce the crystallinity of the starch; increase the compatibility of the starch，wood fiber and PVA, and increase the tensile strength of the composite to 5.91 MPa. Crosslink reaction occurs between PVA molecular under the reaction of cross‐linking agent borax. The tensile strength of the composite increases up to 6.12 MPa. The tensile strength lowers down as a result of the decrease of the contact area caused emerging bubbles with the help of the blowing agent NaHCO3. POLYM. COMPOS., 40:3923–3928, 2019.