Constitutive model research on the hot deformation behavior of Ti6Al4V alloy under wide temperatures

Abstract

The hot compression tests of Ti6Al4V alloy were carried out at different temperatures (500, 600, 700, 800, and 900 °C) and different strain rates (0.01 s−1, 0.1 s−1, and 1 s−1). The hot deformation behavior of Ti6Al4V alloy is complex in a wide temperature range, and the phenomenon of negative strain rate sensitivity even appears at low temperature (500 °C and 600 °C). The phenomenological Arrhenius model, modified Fields-Backofen model (m-FB) and Gray Wolf Optimization Algorithm optimizing BP artificial neural network model (GWO-BPNN) were established. A new FB model (n-FB) was proposed based on m-FB model. The prediction accuracy and computational cost of these constitutive models were evaluated by mean absolute percentage error (MAPE), correlation coefficient (R) and the number of parameters. The comparison results show that the m-FB model, the GWO-BPNN model, and the n-FB model can predict the flow stress at all temperatures, but the Arrhenius model cannot be fitted at low temperature conditions, and the fitting effect of the Arrhenius model is the worst. The GWO-BPNN has the highest prediction accuracy based on that it has the maximum number of parameters. The MAPE and R values of n-FB model are 1.9093% and 0.993 respectively. Compared with the m-FB model, the MAPE value of n-FB is reduced by 64.64%.

Keywords

• Ti6Al4V alloy
• Hot deformation behavior
• Wide temperature range
• Constitutive model
• Artificial neural network