Evaluasi Pemodelan Augmentasi Data Sifat Mekanik Aluminium Menggunakan Generative Adversarial Networks

Leni, Desmarita
Berli, Ade Usra
Kesuma, Dytchia Septi
Haris, Haris
Sumiati, Ruzita

Jurnal Engine: Energi, Manufaktur, dan Material
Vol. 8 No. 1 (2024)

Materials informatics is a new approach in material science that integrates information technology and material science to optimize the discovery of new materials more efficiently and innovatively. In materials informatics, experimental and simulation data are combined with data-driven methods such as big data, data augmentation, and machine learning to gain a deeper understanding of material properties. However, limitations in the availability of samples with desired characteristics and the lack of accurate experimental data pose challenges in materials informatics. In this study, we attempt to address these challenges by modeling the augmentation of mechanical properties of aluminum using Generative Adversarial Networks (GAN). GAN is used to generate synthetic data of aluminum's mechanical properties that closely resemble experimental data. This modeling is trained using experimental testing data consisting of aluminum's mechanical properties and chemical elements in the alloy, obtained from the material database. The dataset comprises 9 chemical element variables in the aluminum alloy and 2 mechanical property variables. The synthetic data generated from the modeling is evaluated using descriptive statistics, Pearson correlation, and Kolmogorov-Smirnov (KS) test to assess the extent to which the synthetic data resembles the original data. The evaluation results indicate that the distribution of synthetic data is similar to the original data. The Pearson correlation results show that most variables of chemical elements and mechanical properties of aluminum in the synthetic data have a correlation that is quite similar to the original data. The KS test results also indicate that the distribution of synthetic data does not significantly differ from the distribution of the original data. This indicates that the synthetic data generated has a high resemblance to the experimental data, enabling its use in materials informatics research. Thus, modeling the augmentation of aluminum's mechanical property data using GAN provides a significant contribution to expanding data availability in material science.