Page 22 - Biaxial Multiaxial Fatigue and Fracture
P. 22

Assessment of Welded Structures by a Structural Multiaxial Fatigue Approach   7


            dispose  at  the  present  time  of  reliable  and  sufficiently accurate methods for  predicting  the
            fatigue  life,  with  regard  to  the  results  obtained  by  modern  finite  elements  methods.  To
            overcome this difficulty, we come back to the structural approach with a clear definition of the
           design stress which can be transposed to multiaxial structural problems.
              The  classical  introduction of  the  design  stress is based  on  the  extrapolation  of  far  field
            stresses with unclear rules. In  order to clarify the description of  this design stress we propose
            an approach deriving from an analogy with the concepts which are at the origin of the Fracture
           Mechanics. It is known that the mechanical state in the highly damaged crack tip zone, called
           by  H.D. Bui  [9,10]  the process zone, is inaccessible by  the usual mechanics of  solids. In  this
            zone, the material is neither really continuous nor  homogeneous and  the local strains are not
            small.  Nevertheless,  the  stress  solution  obtained  from  linear  homogeneous  and  isotropic
           elasticity  in  small  strains  allows the  correct description of  the  mechanical state  outside the
           process zone. Although it is erroneous at the vicinity of the crack tip, it makes sense in terms of
            an  asymutotic  solution  which  allows  the  correct  control  and  the  interpretation  of  the
           phenomena  produced  in  the  process  zone.  Likewise, we  will  look  for  a  way  to  build  the
           asymptotic  solution  which  allows  the  correct  control  and  interpretation of  the  phenomena
           produced  in  the  critical  zone  of  the  weld.  For  that  purpose  we  adopt  an  approach  which
           combines testing and calculations with meshing rules taking into account the local rigidity due
           to  the weld  instead of  the local geometry of  the weld itself  which is a very  hazardous data.
           Therefore,  the  fatigue  design  can  be  based  on  a  structural  stress calculation  from  a  finite
           element analysis. On this basis we can establish design rules for welded structures [ 1 11 with a
           structural approach and an unique S-N curve where S  is a local equivalent stress defined from T
           and p at the Hot Spot as described in the next paragraph.


            Computing procedure and applications

           Welding is one of the most important manufacturing process in the mechanical industry. In the
           present industrial context, engineers need some fast and efficient tools to achieve the design of
           such welded structures. Some computational methods applicable to the prediction of  fatigue
           strength  have  been  proposed  by  different  researchers  [ 1,121.  In  the  automotive  industry,
           J.L. Fayard  et  al.  [ll]  developed  an  efficient  numerical  tool  to  evaluate  the  asymptotic
           mechanical field which defines precisely the design stress state and allows the prediction of the
           fatigue  strength  of  continuous  arc-welded structures. In  most  of  the  cases, components  are
           usually  made  of  metal  sheets of  approximately 2 to  5 mm  thick joined  by  automatic metal
           active  gas  (MAG)  welding.  Thus,  the  thin  shell  theory  was  considered  to  be  the  most
           appropriate  calculation  method  to  solve  the  fatigue  life  prediction  problem  of  automotive
           welded structures.
              However, in a thin  shell finite element model, sheets are described by  their mean  surfaces.
           The outstanding  difficulty  in  using  such meshes lies  in  the  modelling of  the  mean  surface
           intersection. In fact, this zone exhibits 3D behaviour, whereas a thin shell model only produces
           biaxial stresses. Moreover, at the intersection of thin shells, where hot spots commonly appear,
           the stress gradient can be rather steep, so that stress calculations are very sensitive to the mesh
           size. It  is therefore necessary to define a meshing methodology which  can be  systematically
           applied to any welded connection. On this basis, a design rule was established [I 1,13,14]. The
           first idea was to reproduce as precisely as possible the local rigidity induced by the weld to the
           joint,  which  modifies the  local  stress distribution. The other principal idea that  supports the
           meshing strategy was to simulate the stress flow from one sheet to another through the weld.
           For  that  purpose,  rigid  body  elements  were  used  to  link  the  two  shells.  The  <<good  den,
            associated  with  a  design  stress representing faithfully the fatigue phenomenon and  with  an
   17   18   19   20   21   22   23   24   25   26   27