Machine learning-assisted wire arc additive manufacturing and heat input effect on mechanical and corrosion behaviour of 316 L stainless steels

Abstract

Predicting the track forming factor or height-to-width ratio (H/W) in wire arc additive manufacturing is crucial for optimal path planning, heat distribution, structural integrity, distortion control, process efficiency, and defect prevention, ensuring high-quality and reliable components. Different analytical and numerical modeling methods have been introduced to predict the H/W ratio. However, the accuracy of these predictions is relatively low due to the challenges associated with handling complex and non-linear regression equations. This study addressed the challenges by implementing data-driven predictive modeling to predict the H/W ratio from the input process parameters. A stacking ensemble learning approach is employed, where a meta-model integrates predictions from multiple base learners to enhance overall performance. The model performance was evaluated by Coefficient of determination (R2), mean squared error (MSE), and mean absolute error (MAE). The model was validated by comparing the experimental and predicted values based on which five thin walls are printed with different heat inputs. The study also explores the impact of heat input on microstructure, mechanical, and corrosion properties. The findings indicate that a significantly low heat input (178 J/mm) and higher heat inputs (> 356 J/mm) decrease the wire utilization. With increasing heat input, ferrite content increases, microstructures become coarser, dendritic spacing increases, and ferrite transitions from lathy to skeletal. Further, lower heat input (235 J/mm) reduces δ-ferrite, suppresses atomic segregation, and increases Cr and Mo in the matrix. YS and UTS decrease with higher heat inputs, anisotropy decreases, and Vickers microhardness drops from 228 (178 J/mm) to 194 Hv (586 J/mm). Additionally, the corrosion resistance deteriorates with increasing heat input, as evidenced by higher pitting potential (Epit: 0.389 V vs. 0.368 V vs. −0.023 V) and lower corrosion current density (Icorr: 6.76 ×10−7 A/cm2 vs. 7.946 ×10−7 A/cm2 vs. 8.91 ×10−7 A/cm2) for heat inputs of 235 J/mm, 281 J/mm, and 356 J/mm, respectively.