Artificial intelligence-based modeling for predicting precipitates in aluminum alloys

In this study, we tried to predict the precipitation phase that can be precipitated as a stable phase in aluminum matrix using artificial intelligence with simulation database. OQMD (Open Quantum Materials Database) is a database that contains structural properties of vast materials and thermodynamic properties based on DFT calculations. First of all, in OQMD, based on lattice mismatch, search for precipitates that are likely to exist in a stable or metastable phase with the host aluminum lattice at the same time, and extract various factors such as lattice constant, formation energy, and volume per atom to create a basis data for machine learning. Among the precipitates of the refined data, those that could exist as alloys were classified by referring to TCAL8 (TCS Al-based Alloy Database), experimentally proven alloys were selected and classified through machine learning. We tried to implement effective classification performance by modifying the CNN algorithm, which leverages a strong deep structure. And through the results of this machine learning, it is possible to predict new stable precipitation phases that are worth experimentally researching in addition to the previously known aluminum alloys.