About this Abstract |
| Meeting |
MS&T26: Materials Science & Technology
|
| Symposium
|
Grain Boundaries, Interfaces, and Surfaces: Fundamental Structure-Property-Performance Relationships
|
| Presentation Title |
Inverse Prediction of Erosion-Minimizing Parameters for Aluminum Alloys Using ML-Based Optimization |
| Author(s) |
Sheikh Saud |
| On-Site Speaker (Planned) |
Sheikh Saud |
| Abstract Scope |
Water droplet erosion (WDE) significantly affects the service life of aerospace-grade aluminum alloys operating in high-speed environments. This study proposes a Random Forest (RF)-based inverse design framework to identify erosion-minimizing test conditions for Al 2024-T4 and Al 7075-T6. Experimental data were obtained from a rotating-disc WDE rig and used to train RF regression models to predict erosion rate as a function of droplet size, impact velocity, and exposure conditions. The trained surrogate model was integrated with Bayesian optimization to explore the parameter space and identify stable low-erosion regions. Results show that Al 7075 exhibits a broader and more stable low-erosion window (150–300 m/s), whereas Al 2024 demonstrates a narrower stability range (150–217 m/s) with higher sensitivity at increased velocities. The study demonstrates that machine learning can move beyond erosion prediction to optimize test parameters, enabling efficient material evaluation and supporting the design of erosion-resistant aerospace components. |