Page 522 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
        P. 522
     492               SLENDER STRUCTURES AND AXIAL FLOW
                    where r  and B are constants. When  E  # 0, the solution of  (F.42) can still be expressed
                   in  the form of  (F.43) provided that  r  and B  are considered to be  functions of  t  rather
                    than  constants. Since (F.42) and (F.43) constitute two equations for the three variables
                   x, r and B, we have to find an additional equation or impose a constraint condition. For
                   example, it is convenient to assume that the velocity has the same form as for the case
                    when E  = 0, i.e.
                                           i(t) = -r(t)wo  sin(w0t + B(t)).             (F.44)
                      Differentiating  (F.43)  with  respect  to  time  t  and  comparing  the  result  with  (F.44)
                    leads to
                                    i-(t) cos(wot + ~(r)) - rB(t> sin(Wt + ~(t)) 0.     (F.45)
                                                                         =
                    Equation (F.42) is now written in terms of r(t) and P(t) by finding x through differentiation
                    of  (F.44) with respect to t. By using the resulting equation together with the constraint
                    relation (F.45) and solving for i- and b, after some algebra we obtain
                               i- = -(E/wO)f  (r, B) sin $,   rB = -(c/wO)f  (r, B) cos $,   (F.46)
                    where $ = mot + ,t?. For small E,  i. and b are small; this means that r and B vary much
                    more slowly with t  than  $r.  In  other words, r  and #?  hardly change during the period of
                    oscillation 2x/wo of sin l(r and cos $. This enables us to ‘average out’ the variations of
                    $ in (F.46). Averaging these equations over the period 2n/wo and considering r, p, i- and
                      to be constants while performing the integrations, one obtains
                                                                                        (F.47)
                    The full description of the method may be found in, e.g. Nayfeh & Mook (1979), Gucken-
                    heimer & Holmes (1983) or Sanders & Verhulst (1985). One advantage of the averaging
                    method over the normal form method is that it is based on several basic comparison theo-
                    rems  which  compare solutions of  the original equation (F.42) to those of  the averaged
                    equations (F.47). For solutions valid for time of  o(t-’), any  solution of  (F.47) can be
                    shown to be close to those of (F.42) for sufficiently small E. Also, all the qualitative local
                    behaviour of the dynamics of the averaged equations (F.47) corresponds to the same qual-
                    itative and local behaviour of periodic orbits of (F.42). In particular, a stable (unstable)
                    fixed point of (F.47) corresponds to a stable (unstable) limit cycle in (F.42), and a Hopf
                    bifurcation giving rise to an attracting (repelling) limit cycle in  (F.47) corresponds to a
                    bifurcation to a stable (unstable) invariant torus in (F.42), and so on.
                      To  see in practice how the method works, let us consider again equation (F.30). The
                    solution when E  = 0 is simply XI  = r cos(t + B) = rC, x2  = r sin(t + /3)  = rS, where C
                    stands for cos(t + B) and S for sin(t + B). Following the methodology described in the
                    foregoing leads to
     	
