A Prediction Model for the Position of Shrinkage Cavity in Large Steel Ingots

The position of shrinkage cavity in large steel ingots was predicted using artificial neural network (ANN) based on finite element analysis. A finite element (FE) model was established to simulate the casting process of 1.6t Cry steel ingot and FE results were regard as the training samples for ANN model. Pouring temperature, mold filling time and mean diameter of permanent mold, as three influencing factors for the formation of shrinkage cavity., were used as the input of the ANN model. During the net training, genetic algorithm was applied to improve the convergence rate and avoid the net producing local minimization. The ANN results were identical with the data obtained from FE model, which demonstrated the validity of the ANN model. And the optimal ranges of the three influencing factors above were obtained through the analysis of prediction results acquired from the ANN model.