Comments

• kuhl

  March 17

  Hello Jan,
  I'm not sure that I fully understand your model setup and if there can be any local bending in the weld. If there can be, I don't think looking at the sum of the forces within the weld is sufficiently accurate. In a weld modeled with solids I think you would have to extract the nodal forces within the weld cross section and caluculate the bending they cause in that cross section.
  I'm thinking of two flat pieces butwelded together subjected to pure bending normal to the weld. The sum of section forces normal to the weld will be zero, but stress normal to weld will not.
  
  In FKM inside Ansys, weld analysis can be done based on nominal stress approach, similar to eurocode. It is, however, restricted to shell models (I think) and nominal stresses are derived from nodal forces and moments.
  With a solid model, the FKM approach is based on stress linearization towards the weld toe and no longer uses nominal stress.
  
  Maybe a shell model can be an option to reduce complexity for the nominal weld stress calculation.
• JSC

  March 18

  Hi kuhl,
  Thanks for your input, that's certainly one of my concerns. In the picture below a basic model is shown to illustrate my thinking (which definitly can be wrong). Three normal forces are applied at the top of the plate.
  
  
  
  In the pre-defined weld leg areas (blue) the external forces are summed, one-by-one, with the SUM-tool. Any bending moment generated by e.g. Fy appears to cause additions to SUM Fz (setting Fz=0 still gives significant SUM Fz).
  
  
  
  
  Thus having the three resultant forces one can, with a litle trigonometry, calculate the governing stresses (sigma.a / tau.a / tau.p) in the weld throat acc to Eurocode;
  
  
  My basic assumption/understanding is that the SUM-tool extracts and sums up all the nodal forces (x, y & z) in a given area and having these forces is enough to work out the normal and shear stresses in that area. These stresses will be average values over the defined area which is suitable for weld assesment (avoiding peak stresses at root, end and edges). I agree that the method is a lot of labour and post-processing if you have many welds to check but for a limited number it might work.
  /Jan
  
• kuhl

  March 18

  In this example there should be no local bending due to Fy as it is split up into tension / compression in each weld.
  But I think you need to consider that Fx causes bending about the y axis which leads to a linear distribution of Fz and (if Fy and Fz are were zero) therefore to tension at one end and compression at the other end of the weld line while sum Fz would still be zero.
  To capture these tension / compression loads I would say that you would have to divide the weld line into segments and evaluate them individually, but that makes it even more complicated.
  
  As a referecen, here are some hints on how nominal weld stresses can be derived from nodal forces in shell elements depending on the weld type for FKM guideline. Maybe you can adapt this approach for your solid model and eurocode.
  
  Local coordinate definition
  
  
  Both sided fillet
  
  
  One sided fillet
  
  
  Butt weld
  
  
  
  
  "Segment" refers to the segmentation of the weld line into several pieses. For straight weld lines, I think it is recommended to use one segment, for a close loop (pipe welded on plate) 16 segments.
  
  Note this approach considers the forces and moments in the weld, so stress due to bending is "automatically" consideres even if there is only one segment for the entire weld line.
  Also note that nominal stress approach is only applicable if stresses are mainly nominal and that no local stress effects are considerd.
• JSC

  March 18

  Thank you kuhl for your valid pointers. I agree, if Fx>0 there would be a bending moment around y-axis which could be tricky to handle by just summing up the external forces. For a simple example like the above it's probably easier to work out all section forces at the given height and from there determine Mx, My etc and use the elastic modulus for the weld geometry to calculate the various weld stresses.
  
  My initial take on this was to use the following formula and theory for determing the stresses, which applies both for butt welds and fillet welds:
  
  
  
  Where F.Rd is weld capacity, "d" is replaced with "a" for a fillet weld and f.wd is "dimensioning" weld stress from sigma.a and tau.a. Stresses parallell to weld has to be added and the combination checked by suitable interaction formula. Now when I think of it... maybe not the best of ideas, special care must be taken when evaluating sigma.a and tau.a (a=perpendicular to weld length). sigma.a and tau.a will only be equal if alfa=90deg, might skip this and work out sigma.a and tau.a separately from Fy and Fz resp.
  
  The example I am working on is more looking like this. A laterally loaded plate field supported by H-beams. The welds basically there for transfering shear to the deck plate when the beam bends, but there will also be a local bending of the plate in-between the beams which will cause bending- and membrane stress in y-direction. Some vertical stress will pass through the welds due to the loading of the plate.
  
  My view is that there will be very small bedning moment effects as Fx works almost in line with the welds and as do Fy.
  May have to look at a parallell example using shells for the welds.