Page 352 - Biaxial Multiaxial Fatigue and Fracture
P. 352

336                            S. POMMIER

             shear stress than for the Tresca equivalent stress. Nevertheless this increase is not negligible,
             since in tension, the standard deviation for 2,-  is close to 12 % of <   > while it approaches
             15%of<&->inshear.
               Although the scatter is higher in shear than in uniaxial extension, the maximum bound of
             the distribution of the Tresca stress is found to be very similar in these two cases (Table 3). As
             a matter of fact, the boundary conditions, applied on the model of the thin sheet, were not
             adjusted to obtain similar mean values of the mean Tresca equivalent stress or of <Zmax> in the
             two cases, but the same equivalent strain.  Consequently,  with such boundary conditions, the
             higher level of the scatter in shear, as compared with a uniaxial extension, results in a lower
             value of the mean Tresca equivalent stress, and not in an increase of the probabilities for high
             values of the Tresca equivalent stress.


             Table 3. Values calculated at the centre of the grain located at the centre of the model with a
             crystalline orientation as follows: (VI,  yf,  cpz) =(O,O,O).  The standard deviations are calculated
             with  150 random configurations of  its neighbours.  The  model  is  tested  either in  uniaxial
             extension W=O.l% or  in  shear <y  >=0.1%.  The  standard deviations (6) are given as a
             percentage of the mean value of the distribution.



              150       <OW>   s( OW)   <a,>(   1+3& ~q)) <%ax>  s( %ax)  < Zmax’(1+W   Zmax))
              analyses   MPa   (%)    (MPa)          (MPa)  (%)    (MPa)
              Uniaxial   101.7  11.2   135.8         44.2   12.0   60.2
              extension
              Shear     87.6   17.2   132.6          33.5   14.8   48.3



               It  can be concluded fiom these calculations, that with a  similar value of  <T,,&   in  the
             polycrystal, two different loading conditions are not equivalent in terms of the nucleation of
             micro-cracks. As a matter of fact, these calculations show that,  with the same value of <Tmax>,
             the maximum resolved shear stress is significantly higher, within a few grains, in torsion as
             compared with the case of tension.
               However,  the heavily loaded grains are sparse since they are located at the intersections
             between the heavily loaded links associated with each principal direction. Therefore, on the one
             hand, this effect should lead to a reduction of the threshold for crack nucleation in torsion as
             compared with tension, if the threshold is given as a critical value for <T,,,~~>. However, on the
             other hand,  it may also modi@ the conditions for crack coalescence during subsequent crack
             growth. Therefore, it is not easy at this stage, to provide defmite conclusions on the effect of
             the increase of the scatter in an individual grain, on the fatigue life of the entire sample


             Rotating loah
             The last point that is worth to be discussed is the possible effect of the load percolation network
             in the case of rotating loads.
                In  the  first place, the previous results show that for a similar value of <T,,,~~>, there is a
             higher probability for high values of the maximum resolved shear stress under multiaxial than
             under uniaxial loading conditions (Fig. 12). Consequently, even if the mean values of T,,,   are
   347   348   349   350   351   352   353   354   355   356   357