Prediction of static globularization of Ti-17 alloy with starting lamellar microstructure during heat treatment

Isothermal forging experiments of Ti-17 alloy with starting lamellar microstructure were conducted on 2000T hydropress. Materials were deformed to the height reductions of 20%, 40%, 60% and 80% at 820 °C. After deformation, the samples were heat treated for times ranging from 10 min to 8 h at 820 °C, 840 °C and 860 °C. Globularization fraction of alpha phase was obtained by quantitative analysis. On the basis of experimental data, an artificial neural network (ANN) model with a back-propagation learning algorithm was established to predict static globularization kinetics of Ti-17 alloy. The amount of strain prior to heat treatment, heat treatment temperature and time were taken as inputs, and static globularization fraction as output. The results showed that the maximum and mean deviations between the predictions and the experimental data were 3.58% and 1.27%, respectively. The trained neural network had a good performance for static globularization behavior of Ti-17 alloy. A comparison of the predicted value by the neural network and calculated results by the regression method was carried out. The result indicated that the ANN model is more accurate and efficient than the regression method in terms of the prediction of static globularization kinetics of Ti-17 alloy.