Characterization and prediction of streak breakdown using machine learning
Abstract eng: Characteristics of streaks in DNS flow fields of pre-transitional boundary layers are extracted and compared between streaks that induce the formation of turbulent spots via secondary instability and the remainder of the population. The analysis shows that the two classes of streaks differ in a variety of attributes, including their peak amplitude, distance to the wall and size. The data are used in a machinelearning approach to predict transition to turbulence. An artificial neural network identifies the streaks that will induce the formation of turbulent spots. The method is shown to achieve high prediction accuracy at low computational cost which makes the approach suitable for real-time applications.
Publisher:
International Union of Theoretical and Applied Mechanics, 2016
Conference Title:
Conference Title:
24th International Congress of Theoretical and Applied Mechanics
Conference Venue:
Montreal (CA)
Conference Dates:
2016-08-21 / 2016-08-26
Rights:
Text je chráněný podle autorského zákona č. 121/2000 Sb.
Record appears in:
Record created 2016-11-15, last modified 2016-11-15