A comparative study on phenomenological and artificial neural network models for high temperature flow behavior prediction in Ti6Al4V alloy

Abstract

Due to its critical use in lightweight components requiring elevated temperature operation, it is very important to determine and model the high temperature thermomechanical flow behavior of Ti6Al4V. In this study, uniaxial tensile tests were performed at quasi-static strain rates and at temperatures ranging from 500°C to 800°C. The ductile behavior provided at a temperature of 800°C and at a strain rate of 0.001 s -1 can be preferred for forming operations due to the steady state flow behavior. However, stress peaks during deformation at the strain rates of 0.1 s -1 and 0.01 s -1 are indicative of an unsafe zone. For modeling the flow stress behavior, three models including the Artificial Neural Network, Modified Hensel-Spittel and Arrhenius are employed with varying prediction performance as shown by the correlation coefficient (R) and average absolute relative error (AARE) values. Accordingly, the Artificial Neural Network model is claimed to be a more suitable approach for capturing the mechanical behavior of Ti6Al4V within the forming temperature range utilized in this study. • Determination of hot deformation mechanisms and prediction of the high temperature mechanical behavior are vital for hot forming processes. • Modified Hensel-Spittel, Arrhenius and Artificial Neural Network models were used to predict the flow behavior. • The Arrhenius model including the activation energy is more accurate than the modified Hansel-Spittel model. • Neural Network model provides higher accuracy than the other empirical models evaluated in this study.