Deep learning strategies for addressing issues with small datasets in 2D materials research: Microbial Corrosion

Area of Science:

• Materials Science
• Corrosion Science
• Artificial Intelligence

Background:

• Two-dimensional (2D) materials like graphene show promise as protective coatings due to their impermeability and inertness.
• These coatings are effective against both abiotic and microbiologically influenced corrosion (MIC).
• Developing new 2D coatings is accelerated by AI, but limited experimental data hinders model training.

Purpose of the Study:

• To investigate the use of deep generative models, specifically Variational Autoencoder (VAE) and Generative Adversarial Network (GAN), for data augmentation.
• To generate synthetic electrochemical data for expanding small experimental datasets of 2D coatings for MIC resistance.
• To evaluate the performance of VAE and GAN generated data in machine learning models for predicting MIC resistance.

Main Methods:

• Two deep generative models, VAE and GAN, were employed to generate synthetic electrochemical data.
• The models were trained on experimental data from few-layered graphene coatings on copper surfaces.
• The generated synthetic data was used to train and evaluate machine learning models, including neural networks and XGBoost.

Main Results:

• GAN-generated data led to higher accuracy (83-85%) in neural network models compared to VAE-generated data (78-80%).
• VAE-generated data resulted in better performance (90% accuracy) with XGBoost models than GAN-generated data (84-85%).
• Both VAE and GAN synthetic data successfully improved the performance of machine learning models for developing MIC-resistant 2D coatings.

Conclusions:

• Deep learning-based data augmentation using VAE and GAN can effectively expand small experimental datasets for 2D coatings.
• Synthetic data generation is a viable strategy to overcome data limitations in developing AI-driven models for corrosion resistance.
• This approach accelerates the discovery and development of novel 2D materials for advanced protective coatings against MIC.