Page 49 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 49

32                SLENDER STRUCTURES AND AXIAL FLOW

                  is the turbulence kinetic energy per unit mass. In  view of the foregoing, this may also be
                  written as
                                                 co
                                          K  =     E(k)dk = i@ii(k),                  (2.94)

                  in  which E(k) is  the  energy spectrim function,  i.e. the  density of  contributions to  the
                  kinetic energy on the wavenumber magnitude axis (Batchelor 1960).
                    Some  progress  has  been  made  in  understanding the  changing  scales  of  turbulence,
                  as  measured  by  its  spectra  and  expressed in  terms  of  the  scalar  wavenumber k.  The
                  spectra at low  k  (large eddies) often retain  something of  the original unsteady  laminar
                  flow; but, with increasing k, there is a continual stretching of the eddies by  the medium
                  scales, which causes a transfer of  turbulence energy to large k  (small eddies) and  also
                  randomizes the orientation of  the eddies so that turbulence becomes locally isotropic. If
                  the Reynolds number is very  large, the intermediate spectrum is inertial (Le. it sensibly
                  does  not  depend  on  viscosity), and  it  may  be  shown by  dimensional analysis that  the
                  spectrum  is  proportional  to  k-5/3.  For  the  smallest eddies,  where  k > ~(E/V~)'/~, the
                  Kolmogoroff wavenumber, viscosity takes over and causes a decay of the cascading energy
                  with  dissipation rate  E  to  heat. This structure, as described in  the foregoing, enables a
                  dramatic assumption to be made, namely that away from walls, the Reynolds stresses are
                  independent of  urn. In this one respect, turbulent flow may often be easier to analyse than
                  laminar flow.
                    In analysing the boundary layer near walls, the so-called law ofthe wall  is often used. In
                  this discussion, 2-D or axisymmetric boundary layers only are considered. Let UI be the
                  streamwise flow velocity in  the boundary layer and x2  = y  the distance perpendicularly
                  away from the wall. Then, near enough to the wall,  UI = Ul(p, p, U,, y), where U, =
                  (~,,,/p)l/~ the  skin-friction velocity  and  r,,  is  the  shear  stress  at the  wall; thus,  U1
                           is
                  is  independent  of  outer parameters,  such  as  the  overall  boundary-layer thickness,  the
                  free-stream velocity  U, and the pressure gradient when  not too large. Thus,

                                                                                      (2.95)


                  which is the law of the wall. Rotta (1962) predicts the functional form of B by noting that
                  changes in  U1  in most of the region outside the viscous sublayer are independent of  p,
                  because the shear stress is almost entirely due to -pw there. Dimensional analysis then
                  leads to  (y/U,)(aUl/ay) = 1/K 2: 0.41, a universal constant named after von  KBrmBn.
                  After integration, this gives

                                                                                      (2.96)

                  where B = 5.5 for a smooth wall. This proof applies to rough walls,  'fully rough walls'
                  (where p is unimportant even near the wall), and ribletted walls for which there is a drag
                  reduction. The only thing that changes is the value of B, which is lower for rough walls,
                  increasingly with the roughness, and slightly higher for ribletted walls.
                    The law of  the wall has been accepted for the purposes of CFD (Computational Fluid
                  Dynamics), where it often becomes the inner boundary condition, but  it  must be noted
                  that the corresponding law for turbulence intensity is not exactly true when comparing,
                  say, boundary-layer flow and pipe flow; i.e. O/Ur # %(yUr/u).
   44   45   46   47   48   49   50   51   52   53   54