Application of support vector machine in synthesis and processing of butadiene-styrene rubber

Abstract

Prediction models were established for the polymerization and vulcanization process of butadiene-styrene rubber SBR 1712, by applying support vector machine (SVM) and particle swarm optimization algorithms. The results showed that the models with the solid content, Mooney viscosity and molecular weight polydispersity of SBR 1712 as inputs and the amount of initiator, activator and chain transfer agent used in batch emulsion polymerization as outputs produced determination coefficients greater than 0.8 for the training and test sets, and the predicted values from the models were consistent with the experimental values. Another SVM model was developed by using loading amount of furnace carbon black as output parameter and the properties in vulcanization process as input variables, including the shore A hardness, scorch time, minimum elastic torque and minimum viscous torque of SBR 1712. This model had excellent performance in fitting and prediction.