Wear Resistance Design of Laser Cladding Ni-Based Self-Fluxing Alloy Coating Using Machine Learning

To improve the collaborative design of laser cladding Ni-based self-fluxing alloy (SFA) wear-resistant coatings, machine learning methods were applied. A comprehensive database was constructed from the literature, linking alloy composition, processing parameters, testing conditions, and the wear properties of Ni-based SFA coatings. Feature correlation analysis using Pearson's correlation coefficient and feature importance assessment via the random forest (RF) model highlighted the significant impact of C and B elements. The predictive performance of five classical machine learning algorithms was evaluated using metrics such as the squared correlation coefficient (R²) and mean absolute error (MAE). The RF model, which exhibited the best overall performance, was further combined with a genetic algorithm (GA) to optimize both composition and processing parameters collaboratively. This integrated RF-GA optimization system significantly enhanced efficiency and successfully designed multiple composition and process plans. The optimized alloy demonstrated superior wear resistance with an average friction coefficient of only 0.34, attributed to an enhanced solid solution strengthening effect (110 MPa) and increased hard phase content (52%), such as Ni₃Si, CrB, and NbC. These results provide valuable methodological insights and theoretical support for the preparation of laser cladding coatings and enable efficient process optimization for other laser processing applications.