Are the shallow-water equations a good description at Fr=1? Hydrodynamic Instabilities (soon) AIM Workshop, JNCASR Jan 2011 Rama Govindarajan Jawaharlal Nehru Centre Bangalore Work of Ratul Dasgupta and Gaurav Tomar Shallow-water equations (SWE) u t u u u 2 1 p g h h h' h / Gradients of dynamic pressure h ' ' h / 2 Inviscid shallow-water equations (SWE) Pressure: hydrostatic, since long wave u u 1 p (u u ) x g Fr<1 h dh Fr >1 dx u 2 gh C 2 x uh Q Fr 2 u 2 gh Lord Rayleigh, 1914: Across Fr=1 Mass and momentum conserved Energy cannot be conserved If energy decreases, height MUST increase nndb.com U2,h2 U1,h1 Tomas Bohr and many, 2011 ... dcwww.camd.dtu .dk/~tbohr/ The inviscid description Fr 1 U1 Fr 2 1 gH 1 H2 H1 U2 U2 gH 1 2 U1 The transition from Fr > 1 to Fr < 1 cannot happen smoothly There has to be a shock at Fr = 1 c gH 1 Viscous SWE still used at Fr of O(1) and elsewhere In analytical work and simulations 5 Can give realistic height profiles Viscous SWE (vertical averaging) closure problem dh (u u ) x g h dx u yy Fr=1 Singha et al. PRE 2005 similarity assumption parabolic no jump x Better model: Cubic Pohlhausen profile u U ( ) a ( ) b ( ) c ( ) Watanabe et al. 2003, Bonn et al. 2009 2 3 Qf Planar BLSWE u 0 (u u ) x g dh dx u yy Re Q / , Fr Q /( g f + 1/ 2 y h( x) dx , d h ( x ) Re Dasgupta and RG, h 3/2 ) Phys. Fluids 2010 1 2 h ' Re f Re f f f f 2 Fr f ( 0 , ) 0 , f (1, ) 1, f (1, ) 0 In addition f (1, ) 1 EXACT EQUATION: solved as o.d.e. Reynolds scales out h and f from same equation Similarity solutions for Fr >> 1 and Fr << 1 Upstream Watson, Gravityfree (1964) f h ' Re f 0 2 Downstream parabolic profile Velocity-profile does not admit a cubic term Drawback with the Pohlhausen model Although height profiles good 9 BLSWE h ' f f f f f 2 I G VS 1 Fr 2 h' 1 Re f Velocity profiles Low Froude P solution Highly reversed. Very unstable Planar – Height Profile Velocity profile and h’: Functions only of Froude `Jump’ without downstream b.c.! Behaviour changes at Fr ~1 Upstream h ' Re 1 . 814 Circular – No fitting parameter Near-jump region: SWE not good? need simulations of full Navier-Stokes A circular hydraulic jump Simulations Tidal bores Arnside viaduct http://www .geograph.org .uk/photo/3245 81 http://ponce.sds u.edu/pororoca_ photos.html http://www.metro.co.uk/news/article.html?in_arti cle_id=45986&in_page_id=3 The pororoca: up to 4 m high on the Amazon Chanson, Euro. J. Mec B Fluids 2009 Motivation: gravity-free hydraulic jumps (Phys. Rev. Lett., 2007, Mathur et al.) Navier-Stokes simulations – Circular and Planar GERRIS by Stephane Popinet of NIWA, NewZealand Planar Geometry Note: very few earlier simulations Circular: Yokoi et al., Ferreira et al. 2002 Elliptic??? Effect of domain size SWE always too gentle near jump PHJ - Computations Non-hydrostatic effects 23 Typical planar jump U, Fr < 1 J, Fr ~ 1 P, Fr > 1 N, Fr < 1 The story so far I - G + D + B + VS + VO = 0 BLSWE: I - G +VS = 0? Good when Fr > 1.5 Good (with new N solution) when Fr < 0.8 Fr ~1 I ~ G, singular behaviour as in Rayleigh equation (U c )( ' ' ) U ' ' 0 ? 2 KdV: I - G + D = 0 Singular perturbation problem I G D 1 Re h' h / h ' ' ' ' .... WKB ansatz take h’ large 1 / h' / h Lowest order equation O(1) Either is O(R-1) or jump is less singular. With latter Only dispersive terms contribute at the lowest order {Subset of D} = 0 At order {Different subset of D + Vo} = 0 No term from SWE at first two orders Gravity unimportant here!! (Except via asymptotic matching (many options)) h’ need not always be large In fact planar always very weak ~ O(1) or bigger! No reduction of NS Undular region Model of Johnson: Adhoc introduction of a viscous-like term, I-G+D + V1 = 0. Our model for the undular region Conclusions Exact BLSWE works well upstream multiple solutions downstream, N solution works well Behaviour change at Fr=1 for ANY film flow Planar jump weak, undular Different balance of power in the near-jump region gravity unimportant Undular region complicated viscous version of KdV equation Always separates, separation causes jump? ..... Analytical: circular jump less likely to separate Circular jumps of Type 0 and Type II-prime Standard Type I Type ``II-prime’’ Type ``0’’ Circular jump FrN=7.5 Increasing Reynolds, weaker jump Numerical solution: initial momentum flux matters Effect of surface tension Planar jumps – Effect of change of inlet Froude Wave - breaking h' Fr i Steeper jumps with decreasing Fr As in Avedesian et al. 2000, experiment Inviscid: as F increases, h2 increases 2 Planar jumps – Effect of Reynolds 12.5 25 Steeper jumps with decreasing Reynolds 47