Outputting the traction vector on non-contact surfaces

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • danasolav
    Junior Member
    • Jun 2019
    • 8

    Outputting the traction vector on non-contact surfaces

    Hello,

    I want to export the traction vector acting on triangular surfaces (the solid is meshed with tetrahedral elements). These surfaces are NOT contact surfaces
    I couldn't find an option to output this measure directly, so I tried to compute it by multiplying the cauchy stress tensor with the unit vector normal to the surface. However, this traction vector doesn't truly represent the traction on the surface (unless the element is very very small), and I get non-zero tractions for free surfaces.
    Is there another way to output surface tractions?

    Thanks!
  • maas
    Lead Code Developer
    • Nov 2007
    • 3400

    #2
    Hi,

    There is no option to extract traction vectors in FEBio other than for contact surfaces. For surface loads, you essentially prescribe the traction, so you know what it is, and for free surfaces, as you noted, the traction should be zero. The non-zero values you are getting are most likely due to the fact that the plot file only contains an average stress for each element. PostView projects this onto the nodes, but that introduces some interpolation error.
    If you can tell us a bit more about this model, i.e. what loads are applied to this surface and why you want to calculate the traction vector, perhaps we can offer an alternative approach.

    Cheers,

    Steve
    Department of Bioengineering, University of Utah
    Scientific Computing and Imaging institute, University of Utah

    Comment

    • Sebastian
      Junior Member
      • Dec 2019
      • 11

      #3
      Hi Steve,

      Thanks for the response. I could not post messages last week, so Dana helped me post this.

      We are doing a project on prosthetic socket design. We use a method based on FEA optimization. The idea is to use the FEA results to evaluate the normal stress between the limb, liner, and socket system, and to change the socket geometry or material stiffness around the high pressure region to reduce the pressure around that region to get a better load distribution. By iteratively evaluating and changing the prosthetic socket designed, the optimal load distribution could be achieved. By the way, Gibbon is used together with FeBio.

      Previously, we used the normal stress of the interface between the skin and the liner inner surface to evaluate the designed socket. And the Cauchy stress tensors of the elements with one surface coincident with the skin surface were used to estimate the normal stresses of the skin facets (4-node tetrahedral elements are used). On these surfaces, we applied pressure load, and we output the cauchy stress tensors of all the corresponding elements. Because the skin surface is not free surface (the liner covers the skin), this problem could not be found.

      But when we transferred to compute the stress between the liner outer surface and the socket inner surface using the same algorithm (no surface load is applied on these surfaces, except the socket pushing the liner around the lower part from the bottom simulating the patient body weight), we found some stresses around the upper part of liner where there are free surfaces were not zero (both pressure and shear stresses were not zero). After some exploration, we think the reason might be the normal and shear stresses calculated from element Cauchy stress tensor are all at the center of an element (or Guass point), not on the surface of an element.

      We also tried another algorithm, i.e., calculating the nodal stresses of all the nodes on the skin by averaging stresses of all the elements sharing the same node, and averaging the stresses of the three nodes of an triangle to get the final surface stress of one facet of the liner outer surface (Cauchy stress tensor was averaged for each indices). According to the results, the normal stress could be improved a little bit not too much, the shear stress is worse. We are not sure if this algorithm is correct.

      We know the element stress should be a good estimation of the corresponding surface stress, but we will also design a measurement system to measure the pressures of some specific locations and validate the simulation method. So we need the accurate pressure values of specific surfaces, not just estimation values from the corresponding elements. Can you give some suggestions or offer an alternative approach to get the accurate values of the element surface?

      Let me know if it is still unclear.

      Thanks.
      Xingbang

      Comment

      • danasolav
        Junior Member
        • Jun 2019
        • 8

        #4
        Hi Steve,

        Thanks for the answer. It makes sense that projecting the stress from the element to the surface or the nodes yields interpolation errors.
        My follow-up question is:
        We have an interface between solids that we don't represent with contact but rather with shared nodes. If we replace it with a tied contact, I assume we should get an equivalent solution, but will it take much longer to run? and is there a problem if the master and slave have the same mesh densities in this case?
        Or- which way would you suggest to turn our interface into a contact surface (with no sliding) without increasing computation time or complexity too much?

        Thanks,
        Dana

        Comment

        • maas
          Lead Code Developer
          • Nov 2007
          • 3400

          #5
          Hi Dana,

          Why would you want to switch to a contact interface? If the surfaces are conforming and you don't want sliding, then I would avoid the contact interface. If you do want contact, then tied is definitely the way to go, but you will need to spend some time fine-tuning the penalty and other contact parameters to minimize surface separation. In addition, you will be adding more degrees of freedom so FEBio will need to solve larger linear systems. Both the complexity and your run time will definitely increase.

          Cheers,

          Steve
          Department of Bioengineering, University of Utah
          Scientific Computing and Imaging institute, University of Utah

          Comment

          Working...
          X
          😀
          😂
          🥰
          😘
          🤢
          😎
          😞
          😡
          👍
          👎