High-Speed and Stable Deep-Learning Predictions of Anodic Dissolution of Ti-6Al-4V and Ti-4Al-2V Alloys

L Zhao and P Ou and JY Ye and J Rong and XH Yu and YT Niu and L Ma and KY Chen, JOM, 77, 8847-8861 (2025).

DOI: 10.1007/s11837-025-07706-3

Welding Ti-6Al-4V and Ti-4Al-2V alloys can achieve a balance between corrosion resistance and mechanical properties, while offering potential cost advantages. Consequently, elucidating the corrosion behavior at their welded interfaces holds significant engineering value. However, investigating atomic-scale corrosion mechanisms at welded interfaces remains challenging due to the lack of efficient algorithms for large- scale anodic dissolution simulations. We developed a computational framework that integrates the Butler-Volmer model with the deep potential function to simulate anodic dissolution behavior in Ti-6Al-4V/Ti-4Al-2V alloys. Results indicate that the Ti-Al-V deep potential function demonstrates high consistency with density functional theory (DFT) calculations in both Ti-6Al-4V and Ti-4Al-2V systems. Specifically, the root mean square errors (RMSEs) for energy predictions are 2.79 meV/atom and 5.4 meV/atom, with computational speeds 2-3 orders of magnitude faster compared to DFT. Subsequently, the corrosion potential and current density were calculated for Ti-Al-V alloys with varying compositions. In neutral environments, the corrosion current density of Ti-6Al-4V was predicted to be 10-7.657 A/cm2, with experimental measurements yielding 10-7.20 A/cm2. For Ti-4Al-2V, the corresponding values were 10-7.599 A/cm2 and 10-6.80 A/cm2, respectively. The deviations between predicted and experimental values were all within 7%. Overall, this study establishes a new paradigm for the study of anodic dissolution of similar alloys.

Return to Publications page