The background machine learning approaches for wind turbine blade monitoring rely mostly on complete and labeled datasets, which are costly and often impractical to obtain. Additionally, in material science, most methods depend on expensive experimental data, with limited exploration of data augmentation techniques to reduce the cost and effort of model training. The objective of this research addresses these significant gaps by exploring one-class classification for anomaly detection in wind turbine blades and by developing methods to augment existing data for more cost-effective model training in material science. The methods applied in this work for anomaly detection include One Class Support Vector Machine (OCSVM), Support Vector Data Descriptio (SVDD), and Long Short-Term Memory (LSTM) autoencoders. For augmenting experimental material data, noise addition and image manipulation techniques were used. For anomaly detection, the LSTM autoencoder achieved an accuracy of 97.4 percent with approximately 100 percent recall, while OCSVM achieved 89 percent accuracy and 97 percent recall. OCSVM was deemed more suitable due to lower training costs and similar performance. The results for experimental data augmentation showed a 20 percent improvement over previously trained models, with the augmentation technique significantly enhancing performance, especially in models trained with data from different experimental samples. In conclusion, this research demonstrates the effectiveness of one-class classification for anomaly detection in wind turbine blades and highlights the benefits of data augmentation techniques for cost-effective model training in material science.