Collaborative design of composition and process for laser cladding wear-resistant Ni coatings via knowledge-guided graph neural networks

• A novel alloy design framework was developed by integrating a knowledge-guided graph attention network and genetic algorithm. • The integration of complex metallurgical information into the proposed framework achieves 11% higher accuracy. • The designed KG-SFAs feature a multiphase microstructure and exhibit 15% higher wear resistance. • Computational analysis confirms the low hot-crack susceptibility of KG-SFAs and a wide processing range enabling robust high-wear-resistance coating deposition. The increasing extreme-environment wear challenges impose stricter requirements on the wear resistance of laser-cladded Ni-based self-fluxing alloy coatings. This study introduces a novel machine learning framework integrating a knowledge graph with a graph attention network to simultaneously optimize laser cladding processing and Ni-based self-fluxing alloy composition for enhanced wear resistance. By incorporating complex metallurgical process data, the KGAT framework achieves 11% higher accuracy in predicting friction coefficients than conventional machine learning methods. Coupled with a genetic algorithm, this approach designs several new Ni-based self-fluxing alloy coatings, termed KG-SFAs 1–5, which achieve a 15% increase in wear resistance through a multi-precipitate matrix of CrB, Ni 3 B, and M 2 B. Computational analysis confirms the low hot-crack susceptibility of KG-SFAs, predicting minimal defects such as cracks and pores during fabrication, alongside a wide processing window for robust high-wear-resistance coating deposition. The work establishes a generalizable methodology for laser cladding alloy design and provides a transformative framework for optimizing laser-based manufacturing processes.