Study on the Composition Design, Microstructure, Wear and Corrosion Resistant of Duplex Stainless Steels Based on Machine Learning

Abstract

Duplex stainless steels (DSS) had good wear and corrosion resistance, making them potential substitutes instead of martensitic stainless steel as the material for water turbine blades. However, designing a DSS with high wear and corrosion resistance using traditional trial-and-error methods required a significant amount of time and cost. This study proposed a material design method based on machine learning (ML) to accelerate the development of novel DSS. A composition-process-performance database for DSS was established, and four ML model such as K-Nearest Neighbor Regressor (KNR), Ridge Regression (RR), Decision Tree (DT), and Random Forest (RF) were employed to train the database. Predictions of wear and corrosion resistance for DSS were achieved. The predicted and actual values of them demonstrated good consistency. Among the four models, the RF model for microhardness and self-corrosion potential exhibited the best predictive performance with an R² value of 0.90 and 0.87, respectively. Employing the RF model for three rounds of selection obtained three DSS compositions with high wear and corrosion resistance among 69,120 composition-process combinations, then named as 1Cr29Ni11Mo3.5N, 1Cr29Ni8Mo4.5N, and 1Cr29Ni10Mo4.5N. These optimized compositions were further investigated through laser melting deposition (LMD) corresponding samples. Experimental results indicated that the volume ratio of ferrite to austenite in the three samples all reached 3:7. Specifically, 1Cr29Ni11Mo3.5N showed a microhardness of 356 HV_0.2, good wear resistance (1.2579 × 10^–13 m³/Nm of wear rate), and a self-corrosion potential of − 0.12494 V. 1Cr29Ni11Mo3.5N exhibiting high wear and corrosion resistance.

Graphical Abstract