Page 31 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 31

14                SLENDER  STRUCTURES  AND AXIAL FLOW

                  where  6(x - L) is  the  Dirac delta  function; boundary  conditions (2.29) then  reduce  to
                  (2.23). According to Galerkin's method, the solution of (2.30) may be expressed as

                                                          N
                                                                                      (2.31)


                  where the +,  (x) are appropriate comparison fiinctions, i.e. functions in the same domain,
                  9 = [0, L], satisfying all  the boundary  conditions (both geometrical and  naturalt), and
                  qj(t) are  the  generalized coordinates of  the  discretized  system  which  will  eventually
                  emerge by  application of this method (Meirovitch 1967). It is now clear why  it is advan-
                  tageous to recast this problem into the form of  equations (2.30) and (2.23), for it is then
                  possible to use  $j(x)   4,(x), i.e. to use the eigenfunctions given by  (2.27) as suitable
                  comparison functions: suitable, since they satisfy the boundary conditions associated with
                  (2.30), and also convenient, since they are already known.
                    When approximation (2.31) is substituted into the left-hand side of  (2.30), the  result
                  will generally not be zero, but equal to an error function, which may be denoted by %[WN].
                  Galerkin's  method requires that

                                                                                       (2.32)

                  i.e. over the domain, the integrated error, weighted by  +r(x),t should be zero (Finlayson
                  8t  Scriven 1966).
                    Thus, in this example, substituting approximation (2.31) with +,(x)  = 4,(x)  into equa-
                  tion (2.30), multiplying by  &(x)  and integrating over 9 = [0, L], leads to

                      N
                         [EZA')#2irj + [mLGrj + Me&(L)4j(L)]qj] = 0,   r = 1,2, . . . , N,   (2.33)
                      j=l

                  in view of the orthogonality of eigenfunctions (2.27), i.e.
                                             rL


                  where 6,j  is the Kronecker delta (0 for r # j and  1 for r = j). Clearly the system is now
                  discretized. Thus, if  a two-mode approximation (N = 2) is utilized, equation (2.33) may
                  be written in the following matrix form:




                  The  eigenvalues  and  eigenfrequencies  of  this  matrix  system  are  approximations
                  of  the  lowest  two  of  the  continuous  system;  thus,  if  Me = imL,  then  521 =
                  2.018(EZ/n~L~)'/~, = 17. 165(EI/mL4)1/2. The corresponding eigenvectors give, in a
                                   522
                    +Geometrical boundary conditions are of the type w   = 0, while riarural  ones involve forces or moments,
                                                       Ir=O
                  e.g. ~l(a3~/a~3)( = 0.
                              .r=L
                    tThe weighting function comes in  'naturally' if Galerkin's method is derived via variational techniques.
   26   27   28   29   30   31   32   33   34   35   36