Boosting Mechanical Event Classification with Limited Field Data via Conditional GANs and Knowledge Distillation

We propose a two-step approach combining conditional generative adversarial networks and knowledge distillation addressing field-data scarcity. Validated on field data, we outperform baselines, boosting accuracy from $83 \%$ to $99 \%$ with an error rate of $0.6 \%$. © 2024 The Author(s)