Integrated Design of Hardness and Weldability for Multi-Component Fe-Based Hardfacing Alloys Driven by Machine Learning and High-Throughput Computation

Hardness and weldability are critical indicators for evaluating the quality of multi-component Fe-based hardfacing alloys (MFHAs). However, the intrinsic trade-off between these two properties poses a significant challenge to the development of high-performance alloys. Aiming to break the constraints of conventional trial-and-error approaches for the rational design of MFHAs, this study proposes a novel multi-source data-driven strategy. First, a machine learning model was proposed to predict the hardness of MFHAs based on industrial databases and experimental measurements. SHapley Additive exPlanations (SHAP) interpretability analysis was utilized to identify the key features and beneficial elements influencing hardness, which inspired the design of the composition space for high-throughput thermodynamic calculations. Subsequently, surrogate models were developed based on the CALPHAD method within the promising composition space to predict the key welding metallurgical characteristics, including the solidification temperature range and the hot-cracking sensitivity coefficient. Combined with calculations of the carbon equivalent, a multi-objective optimization algorithm based on the Non-dominated Sorting Genetic Algorithm III (NSGA-III) was employed to design a novel high-performance MFHA. The designed alloy achieved a hardness of 58.4±0.9 HRC, with a prediction error within 5% and no macroscopic cracks. Finally, multi-scale characterization techniques were employed to investigate the microstructural mechanisms responsible for the high hardness of the MFHA. This work provides a novel data-driven framework for the integrated optimization of hardness and weldability in the MFHAs.