A WaveNet-based fully stochastic dynamic stall model

Abstract. Accurate modeling of the dynamic stall remains a challenge for design and construction turbine blades helicopter rotors. At same time, wind turbines, instance, are becoming steadily larger, further increasing demands on their structure necessitating even more detailed forces at hand. The primarily used (semi-)empirical models today have long research history invariably based phase-averaged data from oscillating blade pitch experiments. However, much potential accurate uncertainties force peaks is wasted here, since averaging blurs many features response signals. Even computational fluid dynamics can help little in this regard, Reynolds-averaged Navier–Stokes equations practice cannot account cycle variations, scale-resolving require extremely large amounts resources. This paper presents an approach fully stochastic machine learning model that nevertheless simulate these critical properties. Aerodynamic coefficients compared with experimental different test cases. It shown synthetic profiles which be distinguished visually very close to them frequency spectrum generated. Additionally, attention drawn difficulty evaluating such model, as traditional error metrics use. A combination time warping Earth mover's distance provides robust solution problem.