Page 387 - Fluid-Structure Interactions Slender Structure and Axial Flow (Volume 1)
P. 387

PIPES CONVEYING FLUID: NONLINEAR AND CHAOTIC DYNAMICS         363
























                                                 Displacement
              Figure 5.38  The  static  force-displacement  measurements  to  determine  the  stiffness  of  the
              constraining bars  (Figure 5.30) and  the  fitted  smoothened-trilinear model  for  n = 2  and  3  in
                    equation (5.138), labelled ‘quadratic’ and ‘cubic’ (PaYdoussis, Li & Rand  1991a).

              This model may be improved further, to account for the pipe itself being deformable at
              impact (Hunt & Crossley  1975), suggesting that  a more gradual application of  the full
              force following impact is closer to reality.? Hence, ‘smoothened‘ versions of  this model
              are used, as follows:


                  f(V)  = Kn{V - ;(I17  + VbnI  - 1’1 - qbnl)}n,   with   Iz  = 2, 3 or 5;   (5.138)
              least-squares fitting gives the following set of  values: for n = 2, KZ  = 2.7 x  io5, Vb2 =
              0.050; for  Iz = 3, Kj = 5.6 X  lo6, qb3  = 0.044; for  n = 5, K5  = 1.0 X  lo9, qb5  = 0.031.
              These models will henceforth be referred to as ‘quadratic’, ‘cubic’, and ‘quintic’, for short;
              in the quadratic, clearly (q - qb)’  = (q - Ijlb)lq - qbj to preserve functional oddness. The
              resulting approximations to the dynamical restraint stiffness are shown in Figure 5.38 for
              IZ  = 2 and 3; the curve for n = 5  is very close to those shown and is therefore omitted
              to preserve clarity.
                It is found that with this impact model and N 2 3 it is now possible to obtain convergent
              results, while using the correct stiffness K,  and the impact location ct, = 0.65 as in the
              experiments. In all cases [n = 2, 3, 5 in equation (5.138)], the bifurcation diagrams, phase
              portraits, PSDs  and  so on  are qualitatively similar to those  already shown, confirming
              that the route to chaos is via a cascade of period-doubling bifurcations (Paldoussis et al.
              1991a). Of  special interest is the convergence of the various bifurcations with increasing
              N. As  shown in Figure 5.39(a,b), the critical values  of  u for the  Hopf  and first period-
              doubling bifurcations have essentially converged when N  = 4 or 5; for N  > 5, the values
              of  Upd  differ by less than 3.5%.
                Furthermore, the degree of agreement with experiment for some of the key bifurcations,
              as shown in Table 5.4 (left side) is now excellent - discrepancies being of 0(5%), which

              ~~
                ‘Besides, numerical convergence problems continue to arise with the experimentally determined K - O(105).
   382   383   384   385   386   387   388   389   390   391   392