Monitoring Data Augmentation of Spectral Information using VAE and GAN for Soft-Failure Identification

We propose data augmentation of monitoring information using VAE and GAN to reduce the amount of required soft-failure training data. Results show that only 5 samples per failure type are needed for F1-scores above 0.9.