Multiobjective Optimization of Laser Cladding Bainitic Steel Coatings Using Machine Learning and Nondominated Sorting Genetic Algorithm II

Achieving optimal performance in laser cladding bainitic steel coatings is challenging due to the complex interplay between process parameters and properties. This study introduces a multiobjective optimization framework, integrating machine learning (ML) and the Nondominated Sorting Genetic Algorithm II (NSGA‐II), to navigate these relationships. Based on a database of 100 experimental single‐track coatings, six ML models suited for small‐sample analysis are trained. These models establish a predictive relationship between four key inputs (laser power, scanning speed, preheating temperature, and Cu content) and three critical outputs: dilution rate (DR), aspect ratio (AR), and hardness. The best‐performing model, coupled with NSGA‐II, identifies optimal processing windows. SHapley Additive exPlanations (SHAP) analysis reveals that laser power predominantly governs DR, preheating temperature dictates hardness, while scanning speed exerts the strongest influence on AR. A representative Pareto‐optimal solution (1.5 wt% Cu, 800 W, 8 mm s −1 , 147 °C) predicts a coating with a 4.15 AR, 30.2% DR, and 482 HV hardness is obtained. Triplicate experimental validations confirm the model's accuracy, showing a maximum relative error of only 5.96% between predicted and actual values. Moreover, all coatings exhibit no macroscopic defects, with an average hardness of 488 HV, which is approximately three times that of the substrate.