Integrating experiments and phase field method through informatics for tailored corrosion performance of additively manufactured steel microstructures

Additively manufactured super duplex stainless steel (SDSS 2507) is a promising candidate for marine hardware, but its microstructure and passive-film stability are highly sensitive to post-processing. We present an integrated framework that combines experiments, physics-based modeling, and machine learning to predict corrosion performance in laser powder bed fusion (LPBF) of 2507. Coupons were printed and subjected to heat treatments: as-printed, stress-relieved at 400-550°C for 1h, and solution-annealed at 1100°C for 15 min. XRD quantified the microstructure, and the corrosion behavior was evaluated in 3.5 wt.% NaCl using CPP and EIS. Austenite fraction correlated with improved passivity and repassivation. Stress relieving at 500°C/1h gave the best balance of resistance and stability, whereas 550°C/1h caused film breakdown despite high resistance. A neural network trained on electrochemical data predicted key CPP metrics, enabling rapid screening of process-structure-property pathways for seawater-exposed LPBF SDSS 2507.