Page 50 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 50

CONCEPTS, DEFINITIONS AND METHODS                    33

                 In some of the work to be presented later (e.g. in Chapter 7, Volume 2), particular forms
               of the foregoing for pipe flows - containing considerable empirical input - is utilized.
               Thus, for pipe flow, a friction factor, f, is sometimes defined via

                                                                                    (2.97)


               where U is the mean flow velocity; f is given empirically, for instance by  the Colebrook
               equation,
                                     -- - -2  log,o { = -}         ,                (2.98)
                                                         +
                                                            2.51
                                     a
                                      1
                                                     3.7
                                                           Rea
               where Re  is the  Reynolds number  based  on  the  diameter, D, and  k,/D  is  the  relative
               roughness.
                 Reverting now  to  equation (2.85) for a more general analysis of  turbulent flow, it is
               noted that -w is often not measured, but modelled mathematically. For example, by
               means of Boussinesq’s eddy viscosity concept, one may write


                                                                                    (2.99)

               where  K  is  as given  by  (2.93);  v, is  the  eddy  viscosity  which,  unlike  v,,  (or  v = v,,
               in  laminar  flow), it  is  not  a constant but  is  dependent on  the  flow  field. The  chosen
               form depends on the turbulence model adopted - see, for instance, Launder & Spalding
               (1972), Jones & Launder (1972), Launder & Sharma (1974), Rodi (1980), Lesieur (1990),
               So et al. (1991), Wilcox (1993).
                 Perhaps the simplest model is based on Prandtl’s mixing-length hypothesis for 2-D or
               axisymmetric flows, in which

                                              v, = l2                              (2.100)

               where  1 is  Prandtl’s mixing length, y  = x2 is  the  coordinate measured away  from  the
               wall, and  U = U1 is the mean flow velocity.+ In the case of  smooth pipes, for instance,
               Nikuradse’s measurements yield the following empirical expression (Schlichting 1960):
                                 1
                                 - = 0.14-0.08                                     (2.101)
                                 R
               R being the pipe radius.
                 There are many other models, including so-called two-equation models, for turbulent
               flow (Wilcox  1993). One of  the first and  most popular was pioneered by  Launder and
               Spalding. It  is based  on  two  scalar functions, already defined: K  = i-,   the  average
               turbulence (kinetic) energy per unit mass; and E,  the rate of decay of  turbulence energy
               per unit mass, which is also the rate of transfer of energy from the large eddies to smaller
               ones, and hence, in this latter capacity, it is independent of  viscosity. In  this so-called,

                 tIncidentally, this is the equation, with 1  c( y. originally used by  Prandtl to prove the law of the wall.
   45   46   47   48   49   50   51   52   53   54   55