Charilaos Mylonas
During his Ph.D. research, Charilaos worked on machine learning techniques for remaining useful life prediction of wind turbine components and condition monitoring of wind farms. In his research, he has incorporated and extended deep-learning techniques for combining large-scale simulations and real-world condition monitoring data.
Namely, the methodological contributions of his work are on the fusion of Graph Neural Networks with Bayesian Neural Networks for both latent variable models (Variational Auto-encoders for graph-structured data) and discriminative models for graph-structured data (Bayesian Graph Neural Networks). The concrete applications of his research include contributions to the problem of crack detection, the data-driven estimation of the remaining useful life of ball bearings, fatigue estimation of wind turbine blades from coarse monitoring data, and the fully data-driven condition monitoring of wind farms (i.e., generative modeling taking into account turbine-turbine interactions through wake effects).
Currently, Charilaos is working as a Senior Consultant within the Risk Advisory practice of Deloitte AG.