Prediction of the flow stress for 30 MnSi steel using evolutionary least squares support vector machine and mathematical models

To obtain the flow stress data under varying conditions of strain, strain rate and temperature, hot compression experiments are conducted on 30 MnSi steel specimens using a GLEEBLE 1500 thermal simulator. To more accurately predict flow stress, ELS-SVM-MM - the method combining evolutionary least squares-support vector machines (ELS-SVM) with mathematical models is proposed. In ELS-SVM, the optimal parameters for LS-SVM are obtained by particle swarm optimization (PSO). The study represents the application of ESL-SVM-MM in the flow stress prediction. The experiment results have showed that this method can correctly recur to the flow stress in the sample data and it can also predict well the non-sample data. The efficiency and accuracy of the predicted flow stress using the method are better than those with the method combining BP neural networks with mathematical models (BPN-MM). Especially, the generalization performance of the network is improved