Prediction of deformation behaviors based on the neural network algorithm and investigation of dynamic softening mechanisms of TiAl alloys prepared by the spark plasma sintering

Deformation behaviors and microstructure evolution mechanisms of a novel SPS TiAl alloys have been investigated for the design of further powder metallurgy sheet rolling processes. The back-propagation artificial neural network (BPANN) model has been established and employed to modify the Arrhenius constitutive model, by which all data closely conformed to the optimal fitting line and strictly distributed within ±10 % deviation ranges. The processing maps that were calculated by predicted and experimental results indicated the stable processing regions as 1050–1250°C/0.01–0.1 s⁻¹ at low strains (ε=0.1–0.4) and 1200–1250°C/0.001–0.01 s⁻¹ at high strains (ε=0.5–0.6) and represented the unstable conditions as 1150–1180°C/0.001–0.1 s⁻¹. Moreover, the deformation mechanisms of SPS TiAl alloys have been clarified systematacially. The whole deformation was accompanied with dynamic recrystallization (DRX) and dynamic recovery (DRV). The dynamic recovery and continuous dynamic recrystallization of α grains and the discontinuous dynamic recrystallization of γ grains dominate the deformation of SPS TiAl alloys in the (α+γ) dual-phase field. Meanwhile, the deformation in the single α phase field is mainly supported by the dynamic recovery and continuous dynamic recrystallization of α grains. Additionally, phase transformations, twins and special boundaries formation contribute to the grain refinement, which has been directly proven and characterized in this work.