Page 436 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 436

408                SLENDER STRUCTURES AND AXIAL FLOW

                  where A is a matrix, with J and M nonzero submatrices on its diagonal and zero elsewhere;
                  J corresponds to the purely imaginary pair of  eigenvalues, and M to the 2N  - 2 eigen-
                  values with negative real parts, so having sub-elements R,, given as follows:


                                                                                      (5.158)

                  The system is then projected onto a centre manifold, where special care must be exer-
                  cised because  the  system is nonautonomous. To  transform the  system to an  equivalent
                  autonomous one, p sin wt and @  cos WT are replaced by two new variables,

                                                                                      (5.159)

                                         ".Is   0   0  --o   0   0
                  and I*. is included in the system of equations as a trivial dependent variable,   = 0. Thus,
                  J in (5.158)  is replaced by
                                                        0
                                                                 0
                                                            0
                                               0  p2
                                                        0
                                                            0
                                               0   0   P2  0    +  j  j               (5.160)
                                                        ow0
                   the  associated  vector  being  y = {p, vl, 712, XI, ~2)~. The  system  is  then  transformed
                   by  defining  x' =EX,  p'  = ep,  uo - ucf = eq,  and  the  method  of  normal  forms
                   (Appendix F.3)  is  applied:  (a) to  find  all  possible  parametric  resonances  to  O(E) and
                   O(e2), and (b) to determine the simplest set of equations defining these resonances. Three
                   separate sets  of  normal  forms  are  determined:  (i) for  w  away  from  both  2w0 and  WO,
                   (ii) for  w  near  00,  and  (iii) for  w  near  2~0. In  case (ii)  one  obtains  the  fundamental
                   secondary resonance, where the harmonic perturbation in  u appears only at the second
                   order, p2; in the last case, the principal resonance is obtained, where these terms appear
                   to first order, p. The results for the principal resonance are identical to those obtained by
                   Bajaj (1986b).
                     Before  presenting  any  results,  the  second major  component of  this  study is  briefly
                   discussed, namely the solution of the full equations  by numerical techniques. In this case
                   no  restrictions  apply  as to  u0  being  close to  ucf. Three  such  methods are used - see
                   Section 5.4.  (a) The  nonlinear  inertial  terms  are  transformed  into  equivalent  stiffness
                   and  velocity-dependent  terms  [Section 5.2.7(b)],  and  the  resulting  equation  can  then
                   be  integrated  by  a  Runge-Kutta  method;  AUTO  may  also  be  used  in  this  scheme,
                   once  the  equations  are  transformed  into  those  of  an  equivalent  autonomous  system,
                   as  in  equation (5.160).  Solutions  are  also  obtained  by  (b) the Jinite difference  method
                   (FDM) based on Houbolt's fourth-order scheme and (c) the incremental harmonic balance
                   (IHB) method, in both cases with  nonlinear inertial terms intact; a complete expos6 of
                   the  application  of  the  IHB  method  to  the  problem  at  hand  may  be  found  in  Semler
                   et al. (1996).
                     Typical results for the principal resonance are shown in Figures 5.66 and 5.67. Several
                   observations may be made, as follows: (i) it is seen that there is good agreement between
                   the normal-form and numerical solutions for the resonance boundary in Figure 5.66, but
                   less so for the amplitude in Figure 5.67(b), especially away from the resonance boundaries;
   431   432   433   434   435   436   437   438   439   440   441