Construction of Processing Maps based on Expanded Data by BP-ANN and Identification of Optimal Deforming Parameters for Ti-6Al-4V Alloy

摘要

The intrinsic relationships between deforming parameters and microstructural mechanisms for Ti-6Al-4V alloy were analyzed by processing maps. A series of thermal compression tests were carried out in the temperatures range of 1023 similar to 1323 K (across beta-transus) and strain rates range of 0.01 similar to 10 s(-1) on a Gleeble-3500 thermo-mechanical simulator. Based on the stress-strain data collected from compression tests, a back-propagation artificial neural network (BP-ANN) model was developed, which presents reliable performance in tracking and predicting strain-stress data. By utilizing this model, the volume of stress-strain data was expanded. According to the intensive stress-strain data, the apparent activation energy was calculated to be 564.05 kJ mol(-1) and 300.20 kJ mol(-1) for alpha+beta-phase field and single beta-phase field, respectively. Moreover, the processing maps were constructed at finer intervals of temperature, from which, the stable regions with higher power dissipation efficiency (eta > 0.3) and unstable regions with negative instability parameter (xi < 0) were clarified clearly. By combining processing map with microstructure observations, two main stable softening mechanisms, i.e., globularization and dynamic recovery (DRV) were identified, and globularization-predominant (0.3 < eta < 0.55) parameter domain ((epsilon) over dot < 0.1 s(-1)) in alpha+beta-phase field and DRV-predominant (0.25 < (epsilon) over dot < 0.41) parameter domain (0.032 s(-1) < < 1 s(-1)) in beta-phase field were recommended.