Knowledge-aware design of high-strength aviation aluminum alloys via machine learning

Abstract

The development of the aviation industry is accompanied by the continuous research of high-performance aviation aluminum alloys. Stuck in vast untapped composition space and the routine trial-and-error method, efficiently discovering high-strength aluminum alloys remains a significant challenge. To address this issue, we proposed a knowledge-aware design system (KADS) using machine learning (ML) methods to facilitate the rational design of high-strength aviation aluminum alloys. An aviation aluminum alloy database containing 5113 samples was built based on Al–Zn–Mg–Cu, Al–Cu, and Al–Li series aluminum alloys. Notably, guided by the material knowledge, we constructed a feature pool (23 descriptors) to improve the interpretability and accuracy of ML models. Taking key knowledge-aware features as input, we realized the transformation from “element content to property” to “material knowledge to property” in ML modeling, which is the first time proposed in aviation aluminum alloys design. According to the predictive results, we experimentally fabricated a KADS-designed aluminum alloy (KADS-Sc) with superior mechanical strength (812 MPa for ultimate tensile strength and 792 MPa for yield strength). Furthermore, the strengthening mechanisms in KADS-Sc alloy were established quantitatively. The calculations confirmed that the precipitation strengthening (≈ 439 MPa) was most critical in the final strength increment, agreeing with the microstructure analysis.

Keywords

• Machine learning
• Feature engineering
• Aviation aluminum alloys
• High strength
• Strengthening mechanism