About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Corrosion, Protection and Damage Monitoring of Advanced Materials in Natural and Specific Environments
|
| Presentation Title |
Enhancing Corrosion Resistance of Aluminum Alloys Through AI and ML Modeling |
| Author(s) |
Maham Khalid, Farnaz Kaboudvand, Nydia Assaf, Vardaan Sahgal, Jon P. Ruffley, Brian J. McDermott |
| On-Site Speaker (Planned) |
Maham Khalid |
| Abstract Scope |
Corrosion significantly limits the performance of aluminum alloys, particularly in marine and chloride-rich environments. This study applies machine learning (ML) to predict and optimize corrosion resistance using a dataset of over 300 experimental data points compiled from NIST, the Aluminum Alloy Database, and published literature. The dataset includes corrosion rates, alloy compositions, and environmental conditions. We compared three ML models: Random Forest regression, a neural network, and Gaussian Process Regression (GPR), and two approaches were employed: a forward approach to predict corrosion rates from composition and environment, and an inverse approach to identify suitable compositions for a given corrosion rate. Among forward models, GPR and log-transformed GPR performed best, utilizing kernel functions with automatic relevance determination. GPR also demonstrated strong transferability by successfully predicting formation energies of Al-Mg alloys. This work highlights the effectiveness of GPR in modeling complex corrosion behavior and supports broader ML use in accelerated materials design. |