Page 28 - Biaxial Multiaxial Fatigue and Fracture
P. 28

Asse.~sJnent of  Welded Shcctures by a Structural Multiaxial Fatigue Approach   13

            Some remarks on the influence of  residual stresses

            Quantitative analysis of the influence of  residual stresses on a welded structure is in general
            difficult because  the distribution of  these stresses depends on  different factors like  welding
            process, thickness of the sheet, but also on the geometry and the boundary conditions (thermal
            and  mechanical)  of  the  structure.  These  factors  may  modify  the  temperature  history,  the
            rigidity of the thermo-mechanical structure and finally the distribution and the intensity of the
            residual  stresses.  It  is  well known  that  residual stresses can  have  a  great  influence  on  the
            fatigue  behaviour  of  welded  structures  made  of  thick  metal  sheet.  Nevertheless,  as it  has
            already been shown for some automotive body applications, it is most of the time not necessary
            to take the residual stresses into account. The main reason is that the welded sheets are thin so
            that the temperature gradients through the thickness are not so important as in the thick sheet.
            Moreover, the surrounding elements are rather flexible which limits the residual stress level at
            relatively <dong range,  (compared to the thickness of  the sheet and the representative volume
            element). However, a structure like element D represented in Fig. 4 has a high  rigidity due to
            its tubular  shape and,  in  that case, one can expect important residual  stresses. The residual
            stresses have been measured by the X-ray  method, on as welded and after 3000 cycles loaded
            component, in the vicinity of the hot spots of the D elementary structure. No stress relaxation
            was observed as shown in Fig. 8. The value of the principal residual stresses in the vicinity of
            the  hot spots are 250 MPa and -50  MPa; the residual hydrostatic tension is 66 MPa which
            induces a TO decrease of 22 MPa.
              As welded and stress relieved specimens were tested and the results are shown in Fig. 9. On
            one hand, Fig. 9 (a) shows for all the tests results N with respect to the design stress TO for both
            as welded  and  stress relieved elementary structures D.  The value of  q has been  calculated
            without  taking  into  account the  residual stresses. Thus, one obtains  a decrease of  the TO-N
            fatigue resistance curve on the as welded specimens for high cycle fatigue (N 2 lo6 cycles) of
            about 25 MPa which is of the same order of magnitude of the measured stress contribution (25
            MPa compared to 22 MPa). On the other hand, calculations have been performed on the same
            as  welded  elementary  structure D,  considering  the  initial  value  of  the  residual  hydrostatic
           pressure of 66 MPa at the hot spots, as measured experimentally. Figure 9 (b) shows the new
           calculation and test results: while the results for the stress relieved elementary structure D are
            obviously unchanged, the  results for as welded structure increase in  such a way that all  the
           results now fit very well with the mean design curve TO-N. Therefore, taking into account the
            residual stresses gives a much more accurate fatigue life prediction.


           Application of the proposed methods to Sonsino’s results
           The previous computational method was applied to analyse experimental results obtained by
            C.M. Sonsino [I21 . The studied specimen is a tube of  Icm thick welded on a plate of  2.5 cm
            thick, stress relieved and made of  high strength steel (yield stress: 520 MPa). Its geometry is
            presented  in  Fig.  10.  Several  loading conditions were  applied  to  this  specimen  : bending,
            torsion, combined in phase and out of phase torsion and bending.
              The numerical model is built following the recommended meshing rules shown in Fig.  3.
            For in phase loadings (pure bending, pure torsion, in phase bending-torsion), as the principal
            stress directions do not vary, a direct calculation is sufficient. One calculates successively the
            values of  the principal geometrical stress, and then derives Q. For out of phase loadings, as the
            principal stress directions and amplitudes vary at any time of the cycle, it is necessary to apply
           the  general  Dang  Van  procedure.  One  has  to  maximise  the  parameter  d  of  Eq.(4),  that
            quantifies the risk of  fatigue. The maximum occurs at a definite instant t of the loading cycle.
   23   24   25   26   27   28   29   30   31   32   33