Abstract
Hot compression experiments of annealed 7075 Al alloy were performed on TA DIL805D at different temperatures (733, 693, 653, 613 and 573 K) with different strain rates (1.0, 0.1, 0.01 and 0.001 s−1.) Based on experimental data, the strain-compensated Arrhenius model (SCAM) and the back-propagation artificial neural network model (BP-ANN) were constructed for the prediction of the flow stress. The predictive power of the two models was estimated by residual analysis, correlation coefficient (R) and average absolute relative error (AARE). The results reveal that the deformation parameters including strain, strain rate, and temperature have a significant effect on the flow stress of the alloy. Compared with the SCAM model, the flow stress predicted by the BP-ANN model is in better agreement with experimental values. For the BP-ANN model, the maximum residual is only 1 MPa, while it is as high as 8 MPa for the SCAM model. The R and AARE for the SCAM model are 0.9967 and 3.26%, while their values for the BP-ANN model are 0.99998 and 0.18%, respectively. All these reflect that the BP-ANN model has more accurate prediction ability than the SCAM model, which can be applied to predict the flow stress of the alloy under high temperature deformation.
Highlights
It should be known that the average absolute relative error (AARE) obtained by the artificial neural network model is almost less than 1%, which indicated that the ANN model had high prediction accuracy of the flow stress for aluminum alloys
The results reveal that the stress predicted by the strain-compensated Arrhenius model (SCAM) model is less consistent with the test values than that predicted by the back-propagation artificial neural network model (BP-ANN) model
For predicting its deformation behavior, constitutive equations based on SCAM model and BP-ANN model were established, and the prediction ability of the two models was evaluated and compared
Summary
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. The flow behavior of the alloy is nonlinear, so the linear fitting method used in the establishment of the SCAM model is easy to produce a fitting error, and leads the constitutive model obtained to be not ideal. Luo et al [28], established a BP-ANN model to predict the thermal deformation behavior of the 7055 alloy, and confirmed that the model has a high prediction accuracy, with an average relative error of only 0.813%. It should be known that the AARE obtained by the artificial neural network model is almost less than 1%, which indicated that the ANN model had high prediction accuracy of the flow stress for aluminum alloys. The predictive power of the two models for flow stress was evaluated and compared by residual analysis, correlation analysis and prediction error analysis
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
