Abstract: Amorphous alloy coatings are emerging as very promising corrosion-resistant materials for surface engineering applications and their corrosion resistance is greatly sensitive to porosity feature, which is dependent upon the process parameters during thermal spraying. However, the preparation of high-quality thermally sprayed coatings with controlled porosity has been an ever-present challenge. In this paper, based on high-throughput preparation, a self-built dataset containing 336 coating samples was established for machine learning algorithm optimization to obtain Fe-based amorphous coatings with low porosity and high corrosion resistance. Three advanced automated machine learning algorithms: AutoGluon, FLAML, and TPOT, were employed to identify the intricate relationship between high-dimensional process parameters and the two target properties. Accordingly, two high-quality prediction models for porosity and corrosion day were developed with R² values of 0.91 and 0.98, respectively. Further, based on a well-trained porosity model, an interpolation space containing 11,039 spraying process parameters was created. The optimal coatings predicted in this parameter space achieve less than 0.5% porosity and exhibit excellent long-term corrosion resistance, including stable high-impedance modulus values over 50 days and no rusting after 100 days of immersion testing. Additionally, the attainment of optimal coatings is attributed to the reduction of splattering during deposition and higher than 75% proportion of disk-like splats. This research strategy enables the controlled preparation of thermal spray coatings with different porosities and provides innovative ideas for the efficient and reliable preparation of high-performance thermal spray coatings.

Key words: Amorphous alloy coating, Corrosion protection, Machine learning, Thermal spraying, Spraying process optimization