The hot deformation behavior of the aerospace Ti-10-2-3 alloy was investigated by isothermal compression tests at temperatures of 740 to 820 • C and strain rates of 0.0005 to 10 s −1 . The results show that the studied alloy is extremely sensitive to deformation parameters, like the temperature and strain rate. The temperature mainly affects the magnitude of flow stress at larger strains, while the strain rate not only affects the value of flow stress but also the shape of the flow curves. At low strain rates, the flow stress increases with strain, followed by a broad peak and then remains almost constant. At high strain rates, the flow curves exhibit a hardening to a sharp peak at small strains, followed by a rapid dropping to a plateau caused by dynamic softening. In order to describe such flow behavior, a constitutive model considering the effect of deformation parameters was developed as an extension of an existing constitutive model. The modified constitutive model (MC) was obtained based on the original constitutive model (OC) by introducing a new parameter to compensate for the error between the experimental data and predicted values. Compared to the original model, the developed model provides a better description of the flow behavior of Ti-10-2-3 alloy at elevated temperatures over the specified deformation domain.Metals 2019, 9, 844 2 of 17 represented by the sine-hyperbolic law of the Arrhenius model, has been widely used to describe the hot deformation behavior of metallic materials. Lin et al. [10] proposed a modified Arrhenius model to characterize the correlations between the flow stress, strain rate, and temperature of 42CrMo steel at high temperatures through the compensation of strain and the strain rate. Some other phenomenological models, like Johnson-Cook (JC), modified Johnson-Cook (MJC), modified Zerilli-Armstrong (MZA), Fields Backofen (FB), and modified Fields Backofen (MFB), have successfully been used to describe the hot flow behaviors of 20CrMo alloy [11], Ti-6Al-4V alloy [12], Ti-6Cr-5Mo-5V-4Al alloy [13], AZ31 magnesium alloy [14], and AZ31B magnesium alloy [15], respectively. The above-mentioned models are widely used for predicting flow stress at the post-peak stage but are seldom applied to express the flow behavior at the prior peak region, especially for flow curves with a sharp peak at a small strain. In this case, physically-based models and ANN models are considered as effective methods to model the flow curve with a sharp peak. Zhang et al. [16] set up a physically-based constitutive model to describe the flow behavior of a high strength aluminum alloy in hot working conditions. The Bammann-Chiesa-Johnson (BCJ) [17] and mechanical threshold stress (MTS) [18] models are also applied to deformed metallic materials with a satisfactory accuracy. The limitation of physically-based models is that they need more parameters, which have to be calculated from the tested data through a complex algorithm. ANN models based on biological neural networks have been widely applied for the pr...