第１部門①

[2A01-06-02] Corroded Surface Prediction of Weathering Steel Plates under Different Corrosive Environments Using Generative Adversarial Network

*Feng Jiang¹, Mikihito Hirohata¹ (1. Osaka University)

Keywords:Steel, corrosion, machine learning, GAN

In this research, two kinds of weathering steel plates were subjected to four corrosive environments for different time periods. A system based on GAN data augmentation with UNet as the generator and MobileNetV2 as the discriminator was built. The system can simulate three stages of corrosion based on the dataset collected from experiments. It can also predict the corroded surface of steel plates in the next stage. The discriminator of the system can be used to classify the type of weathering steel and days of corrosion. Based on comparative experiments, our system exhibits outstanding performance and outperforms the baseline model.