Generative multi-domain transfer learning for fault detection in data-scarce wind turbines
2026-08-31 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address the problem of detecting faults in wind turbines when there is very little fault-free data to learn from, which happens with new turbines. They use a special type of neural network called StarGAN to transform data from turbines with little data so it looks like data from turbines that have lots of data. This helps existing models trained on rich data to detect faults better in poor-data situations. Their method works well even with less than two weeks of training data and is better than other ways like simple fine-tuning. They also suggest a way to check if the model might do badly, even without knowing about faults beforehand.
wind turbinesfault detectionunsupervised anomaly detectionSCADA dataStarGANdomain mappingdata scarcitynormal behavior modelensemble fusionproxy metric
Authors
Stefan Jonas, Angela Meyer
Abstract
Normal behavior models have shown promise for reliable fault detection in wind turbines. However, these unsupervised anomaly detection models require sufficient fault-free training data to learn the normal operation behavior of turbines. Under data scarcity, for example in newly deployed wind turbines, these models may result in poor fault detection performance. In this work, we propose a multi-domain generative domain mapping approach based on Star Generative Adversarial Networks (StarGAN) to improve fault detection on data-scarce wind turbines. Our model maps SCADA measurements from a data-scarce turbine to resemble those of several data-rich turbines. By preserving the operational state during translation, faults occurring in a data-scarce domain can be mapped and detected by reliable pre-trained normal behavior models of data-rich domains. Highlighting the benefits of an ensemble fusion strategy, we show that under severe data scarcity our method can produce anomaly scores comparable to models trained on large representative datasets. Our approach can consistently outperform models trained on scarce data when less than 2 weeks of training data are available. With just 2 weeks of accumulated training data, we achieve an anomaly score similarity that is, on average, +16% higher than conventional fine-tuning, and +10% higher than single-source domain mapping. As a step towards unsupervised model selection, we propose a proxy metric that detects poor performance at training time, despite an absence of anomalies. Our study presents the potential and challenges of multi-domain mapping for wind turbine fault detection under unrepresentative training data.