Page 343 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 343

PIPES CONVEYING FLUID: NONLINEAR AND CHAOTIC DYNAMICS         323

              global branches of some of the orbits, and towers, which are sequences of period-doubling
              and saddle-node bifurcations. Champneys goes on to show the existence of chaotic regions
              in this and in a subsequent paper (Champneys 1993) where the system asymmetry was
              removed - as discussed in Section 5.8.5.

              fb)  3-0 motions
              A very sophisticated analysis of three-dimensional motions of the N  = 2 system - effect-
              ively  generalizing  to  3-D  the  foregoing  analysis,  but  utilizing  entirely  different
              techniques - has been conducted by Bajaj & Sethna (1982a,b). The equations of motion
              used  are  equations  (5.74)-(5.77);  hence  the  flow  velocity is  motion-independent. The
              springs are considered to be so designed as to allow both planar motions in two directions
              and  rotational  motion  with  zero  torsional  stiffness - a  challenging design  problem if
              attempted experimentally.
                The particular problem investigated is the loss of  stability by  flutter. Because of  the
              rotational symmetry of  the system, a double pair of complex eigenvalues crosses simul-
              taneously  the  imaginary  axis  from  negative  to  positive.  The  nonlinear phenomena  in
              this  case  are  more  complicated  than  those  associated  with  simple  Hopf  bifurcations
              (Figure 2.1 1); e.g.  supercritical bifurcations  do  not  necessarily  imply  a  stable  system
              in this case (Iooss & Joseph  1980).
                After considering the linear dynamics, the problem is transformed into Jordan canonical
              form. Then,  periodic  solutions of  the  nonlinear equations are  analysed by  the  method
              of  alternate problems  (Hale  1969; Bajaj  1981,  1982), which  is  similar in  spirit to  the
              Lyapunov-Schmidt method (Appendix F, Sections F.6.2 and F.6.3). Two independent sets
              of  periodic  solutions are  found to exist: clockwise or counterclockwise rotary motions
              about the x-axis and planar transverse motions. Their stability is determined by the Floquet
              exponents of  the corresponding variational equations, leading finally to the following set
              of interesting results:

                 (i)  both supercritical and subcritical solutions of both the rotary and planar kinds are
                     generally possible for 0 < 3 < 3  (0 < #I < 1) and for given ranges of  a, K  and
                     y, as defined by  (5.68); as already mentioned, these are associated with double
                     pairs of eigenvalues crossing the imaginary axis;
                (ii)   if  both planar and rotary motions are supercritical, then the one with the larger
                     amplitude is stable and the other unstable, whereas normally one would expect
                     all supercritical solutions to be stable;
                (iii)   for  a  given p, if  either  of  the  solutions (rotary or planar)  is  subcritical, both
                     solutions are unstable.

                Typical results are shown in Figure 5.15, where  l/lp201'/~ is a measure of  the ampli-
              tude. In Figure 5.15(a) it is seen that both rotary and planar supercritical solutions exist.
              For 8 < 0.5 1,  the latter being larger, the planar oscillations are stable and hence should
              physically materialize; for B > 0.51, it is the rotary motions that are stable. When gravity
              is present, as in  Figure 5.15(b), the  situation is more complex. Thus,  for p < 0.45, we
              once again have stable planar motions, while for 0.45 < B < 1.19 we have stable rotary
              motions.  For p > 1.19, however,  there  also  exist  subcritical  solutions,  not  present  in
              Figure 5.15(a);  since  one  of  the  solutions is  subcritical, both  are  unstable. This  does
              not imply that there are no stable periodic solutions for   > 1.19; it merely reflects the
   338   339   340   341   342   343   344   345   346   347   348