Artificial Neural Network Modeling to Evaluate the Dynamic Flow Stress of 7050 Aluminum Alloy

The flow stress data have been obtained by a set of isothermal hot compression tests, which were carried out in the temperature range of 573-723 K and strain rates of 0.01, 0.1, 1, and 10 s⁻¹ with a reduction of 60% on a Gleeble-1500 thermo-mechanical simulator. On the basis of the experimental data, constitutive equation and an artificial neural network model were developed for the analysis and simulation of the flow behavior of the 7050 aluminum alloy. After training with standard back-propagation learning algorithm, the artificial neural network model has the ability to present the intrinsic relationship between the flow stress and the processing variables. In the present model, the temperature, strain, and strain rate were chosen as inputs, and the flow stress was chosen as output. By comparing the values of correlation coefficient and average absolute relative error, the prediction accuracy of the model and the improved Arrhenius-type model can be evaluated. The results indicated that the well-trained artificial neural network model is more accurate than the improved Arrhenius-type model in predicting the hot compressive behavior of the as-extruded 7050 aluminum alloy. Based on the predicted stress data and experimental stress data, the 3D continuous stress-strain maps at different strains, temperatures, and strain rates were plotted subsequently. Besides, the flow stress values at arbitrary temperature, strain rate, and strain are explicit on the 3D continuous stress-strain maps, which would be beneficial to articulate working processes more validly.