General Mechanical

General Mechanical

Topics relate to Mechanical Enterprise, Motion, Additive Print and more

GPAD adaptative re-meshing not doing anything

    • Carles Domenech

      Hello, first post here, hope I'm not breaking any rule.

      I've tried to use the new geometry-based adaptivity feature, but it seems i does nothing.

      I'm running a Static Structural analysis with a simple solid part, one force, and a fixed support.

      I haven't specified any mesh settings (thus it's quite coarse).



      After running the simulation, the mesh doesn't change, it's still coarse and quite bad. The simulation time is not split into substeps (with different meshes).

      What am I doing wrong? Also, what's the energy coefficient?


      Thank you

    • mjmiddle
      Ansys Employee

      The adaptvivity doesn't do anything when:

      1. The time occurence criteria is not met, such as determined by the equally spaced points value.
      2. The adpativity criterion is not met, such as the energy coefficient you have specified

      Check the solution information. If there is no section where it checks the energy criterion, reporting the model's value and the user-specified value, then if wasn't triggered by the time occurence criterion. If it did meet the time occurence, but your model has criterion above the threshold set, then it won't do anything. It seems that you think the adpative region should determine the time of substeps by your sentence: "The simulation time is not split into substeps (with different meshes)." But it does not do this. Your model looks linear, and the documentation for it states the large deflection must be off. So if you have not specified multiple substeps, or an time steps less that the end time in the Analysis Settings, it's going to do only one substep and this will be at the endtime (probably at 1 sec). So the geometry based adaptivity will only check once at the end of the analysis.

      The nonlinear adaptive region is similar, so you'll probably want to look at that documentation also:

      Here as my writeup of the time checking:

      Values for "Check At":
      1. "Equally Spaced Points"
      Divides time range by the "Value"
      "Time Range" is either:
          1. "Entire Load Step"
              It takes the time range of the entire load step and divides by the "Value" for the "Check At" and checks at time increments of this time. Example:
              Load step is 0-1sec
              "Value" is 5
              1/5=0.2sec, so it checks every 0.2 sec: 0.2, 0.4, 0.6, 0.8, 1.0sec.
          2. "Manual"
              Specifies the time range in which to divide the number of "Value."
              So for a load step 0-1sec, you set range 0.75-1sec, and with Value of 5, it does:
              0.25/5=0.05sec, so it checks every 0.05sec after 0.75sec: 0.75, 0.8, 0.85, 0.9, 0.95, 1.0sec

      2. "Specified Recurrence Rate"
      Checks every time the number of substeps done increases by the "Value"
      "Time Range" is either:
          1. "Entire Load Step"
              Checks wihin the entire load step, every n substeps, where n="Value"
          2. "Manual"
              Checks withing the specified time range every n substeps, where n="Value".

      I usually use the mesh criterion for a nonlinear adaptive region, but the geometry based adaptivity doesn't allow that criterion (onlt energy and box). For understanding the energy criterion, the documentation page has some links to the APDL documentation for the section on nonlinear mesh adaptivity as well as the command NLADAPTIVE,,,ENERGY:

      Refinement via general remeshing is supported with energy-based (NLADAPTIVE,,,ENERGY), or position-based (NLADAPTIVE,,,BOX) criteria only.

      2.5.1. Energy-Based Criterion

      This criterion is defined for current-technology structural 2D and 3D solid elements. It is based on the magnitude of strain energy of the element compared to the mean strain energy of components to which the element belongs. During the substep at which nonlinear mesh adaptivity criteria are checked, if the element’s strain energy is greater than or equal to the mean strain energy of its components times the user-defined VAL1 (NLADAPTIVE,,ADD), the element is either refined via general remeshing or it is split (NLMESH,REFA,SPLIT), depending on the element types.

      This criterion is used to refine the mesh to achieve high-accuracy simulation in regions where a high concentration of stress exists and elements are too large. It can also be used to refine the mesh at certain intervals of substeps if a very small value or 0 is input for VAL1.

      • Carles Domenech

        I've split the simulation into 4 substeps and now it seems to work as expected. I did believe the Geometry Based Adaptivity object would do the split for me.


        Thank you

Viewing 1 reply thread
  • You must be logged in to reply to this topic.