Thermal deformation behavior of Mg–3Sn–1Mn alloy based on constitutive relation model and artificial neural network

The flow stress behavior of Mg–3Sn–1Mn alloy during thermal compression deformation was systematically studied. The thermal compression simulation experiment was carried out at different deformation temperatures and different strain rates in the range of 523–673 K and 0.001-1s−1, respectively. It is found that at low temperatures and high strain rates, a large number of twins were generated at the initial stage of thermal deformation, causing an increase in the corresponding flow stress, which makes the traditional constitutive relation model insensitive to predicting the thermal deformation behavior of Magnesium (Mg) alloys with twinning effect. To better evaluate the rheological behavior of Mg alloys, an artificial neural network (ANN) model based on a feedforward and back-propagation algorithm was developed to predict the thermal deformation behavior of Mg–3Sn–1Mn alloy affected by twinning phenomena. The inputs of the model were deformation temperature, strain rate and strain, the output was the flow stress. The comparative evaluation of the obtained results using statistical standard R2 and relative error R¯. The correlation coefficients predicted by the constitutive model were 0.964 and 0.869 at low and high stress levels, respectively. And the correlation coefficient of the neural network predictive model was 0.992. The result shows that the trained ANN is more accurate than the traditional constitutive relation model in predicting the thermal deformation behavior with the twinning effect.