Comparative Analysis on the Phenomenological and Artificial Neural Network Modeling for Flow Curves of a Beta Titanium Alloy
Abstract
This study investigates the elevated temperature mechanical behavior of Ti-5V-5Mo-5Cr-4Al alloy through uniaxial tensile experiments conducted at temperatures ranging from room temperature to 550°C and strain rates of 0.001, 0.01, and 0.1 s−1. The results reveal that the dominant softening mechanism is dynamic recovery, whereas dynamic precipitation took place at the lowest rate of deformation and at temperatures ranging from 400°C to 500°C. To predict the mechanical behavior of this recent beta titanium alloy, artificial neural network (ANN) approach and modified Hensel-Spittel (m-HS) model were employed. In the prediction of flow curves using the m-HS model, a correlation coefficient (R) of 0.901 and an average absolute relative error (AARE) of 8.891 % were obtained. In contrast, the ANN approach yielded significantly better results, with an R value of 0.997 and an AARE of 2.3 %. The findings from this study provide routes for determining the hot workability of next-generation metastable beta titanium alloys.