Generative adversarial networks for labeled acceleration data augmentation for structural damage detection

There has been a major advance in the field of Data Science last few decades, and these have utilized for different engineering disciplines applications. Artificial Intelligence (AI), Machine Learning (ML) Deep (DL) algorithms civil Structural Health Monitoring (SHM) especially damage detection applications using sensor data. Although ML DL methods show superior learning skills complex data structures, they require plenty training. However, SHM, collection from structures can be expensive time taking; particularly getting useful (damage associated data) challenging. The objective this study is to address scarcity problem This paper employs 1-D Wasserstein Convolutional Generative Adversarial Networks Gradient Penalty (1-D WDCGAN-GP) synthetic labelled acceleration generation. Then, generated augmented with varying ratios training dataset Neural Network DCNN) application. results that WDCGAN-GP successfully tackle vibration-based structures. Keywords: (SHM), Damage Detection, DCNN), GAN), (WGAN-GP)