Data-enabled prediction of streak breakdown in pressure-gradient boundary layers

• Published in: Journal of fluid mechanics Vol. 801; pp. 43 - 64
• Main Authors: Hack, M. J. Philipp, Zaki, Tamer A.
• Format: Journal Article
• Language: English
• Published: Cambridge, UK Cambridge University Press 25.08.2016
• Subjects: Amplitude; Artificial neural networks; Boundary layer; Boundary layer transition; Boundary layers; Breakdown; Computational fluid dynamics; Computer applications; Computer simulation; Direct numerical simulation; Fluids; Information storage; Instability; Mathematical models; Mechanical engineering; Momentum; Neural networks; Pressure; Pressure gradients; Probability density functions; Probability theory; Properties; Simulation; Spots; Stability; Studies; Transportation networks; Turbulence; Velocity; Vortices; boundary layer stability; transition to turbulence; instability
• DOI: 10.1017/jfm.2016.441
• ISSN: 0022-1120

Streaks in pre-transitional boundary layers are analysed and their properties are extracted from direct numerical simulation data. Streaks that induce breakdown to turbulence via secondary instability are shown to differ from the remainder of the population in various attributes. Conditionally averaged flow fields establish that they are situated farther away from the wall, and have a larger cross-sectional area and higher peak amplitude. The analysis also shows that the momentum thickness acts as a similarity parameter for the properties of the streaks. Probability density functions of the streak amplitude, area, and shear along the streaks, collapse among the various pressure gradients when plotted as a function of the momentum thickness. A prediction scheme for laminar–turbulent transition based on artificial neural networks is presented, which can identify the streaks that will eventually induce the formation of turbulent spots. In comparison to linear stability theory, the approach achieves a higher prediction accuracy at considerably lower computational cost.