Page 261 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 261

242               SLENDER STRUCTURES AND AXIAL FLOW

                     of  u,d  versus  E,+  while  keeping  A  constant  and  infinite - what  might  be  termed  a
                     Euler-Bernoulli  refined-flow theory, which of  course is physically impossible since the
                     refined-flow effects  come  into  play  for  small  E,  when  Euler-Bernoulli  theory  is  not
                     applicable. These results for incompressible flow  (Mach number  = 0) show that,  as  E
                     is decreased, u,d  is  raised, because of  the reduced effective virtual mass - contrary to
                     the results in Figure 4.19, where both A  and E  are varied together.
                       The  above discussion is essential in  understanding the  results presented by  Johnson
                     etal., the most  notable of  which are the following: decreasing the slenderness E  while
                     keeping  A  constant  (i) raises  u,d  for  subsonic  flows, and  (ii) lowers  it  for  supersonic
                     flows: also, reducing the sonic speed always diminishes u,d.
                       Whereas the results obtained for low Mach numbers are probably sound, this is ques-
                     tionable in  the case of near-sonic and supersonic, indeed hypersonic, flows because, as
                     admitted by  the  authors, there  are  fundamental weaknesses in  the  model  used,  which
                     supposes the fluid flow to be wholly isentropic. In the case of compressible flow, there
                     are  ‘secondary effects’ of  fluid friction which generally cannot be ignored, e.g. causing
                     choking: also, for the oscillating pipe,  shock waves are generally inevitable. These real
                     effects, which  are difficult to  model in  a  simple way,  are not  accounted for  and  their
                     influence on the dynamics is unknown.


                     4.5  PIPES WITH HARMONICALLY PERTURBED FLOW

                     In all of the foregoing, except in the derivation of the equations of motion in Section 3.3,
                     the  mean  flow has been  taken to  be  steady  (zi  = 0). Here, the  case  of  a harmonically
                     perturbed flow is considered; i.e. it is supposed that a time-dependent harmonic component
                     is superposed on the steady flow, such that

                                                 u = uo(1 + /1.  cos wt),                 (4.69)

                     where  p is  generally  small.  This  form  of  u  may  induce  another  type  of  instability,
                     namely oscillations due to parametric resonances. These are akin to the parametric reso-
                     nances experienced by, say, a pinned-pinned  column subjected to a compressive end-load,
                     F  = Fo(1 + p cos wt). Especially in the case of  W/WI = 2, where w1  is the first-mode
                     natural frequency of the column, it is easy to appreciate physically that F  pushes down
                     most when the column end moves in the same direction ‘naturally’, at the two extremes
                     of the oscillation cycle; hence, work is done on the column, resulting in amplification of
                     the motion, i.e. in a parametric resonance (Bolotin 1964: Evan-Iwanowski 1976; Schmidt
                     & Tondl  1986). Clearly, w/w1 = 1 also leads to parametric resonance, although, as will
                     be  seen, other frequency ratios can also give rise to resonances. What renders the pipe
                     conveying fluid particularly interesting and worth studying are the differences in dynam-
                     ical behaviour vis-&vis  the column problem, because of the presence of gyroscopic terms
                     and the fact that, in the case of  a cantilevered pipe, the system is nonconservative.
                       The first to consider the problem, in terms of pressure pulsations arising from a pump,
                     for example, was Roth (1964) and, in terms of a pulsating flow velocity, Chen (1971b)
                     and Chen  & Rosenberg (1971). However, as discussed in  Section 3.3.2  in  conjunction
                     with equation (3.38), one of  the terms in Chen’s equations of motion is in error.


                       +In fact the results are presented in terms of the  ‘aspect ratio’ a/L, the inverse of the slenderness E.
   256   257   258   259   260   261   262   263   264   265   266