| Abstract Scope |
Artificial intelligence technologies have significantly advanced the development of materials science, however, their application and exploration in the field of additive manufacturing superalloys remain relatively limited. Herein, we propose a set of methodologies for sample preparation, composition analysis, and defect characterization of additive manufacturing superalloys, which are designed to collect experimental data (>1300 data points) related to such alloys. By training machine learning models on the acquired data, a high-precision defect prediction model for additive manufacturing superalloys is established (R2>90%). Machine learning models were employed to evaluate the crack susceptibility of typical alloys, namely ABD-850AM, ABD-900AM, IN718, IN738LC, Haynes 282, CM247LC and other alloys, and the obtained evaluation results can accurately predict the cracking tendency of these alloys. Combined with interpretability analysis, the contribution of major elements to the alloy’s crack susceptibility was determined, which is expected to provide guidance for the rational design of high-performance superalloys for additive manufacturing. |