Understanding multistep analyses

Tips & Tricks
1

Recently, there has been quite a lot of confusion around multistep analyses so let’s try to summarize the important rules (based on Abaqus but it should be the same in CalculiX), assuming that we are talking only about non-perturbation procedures:

  1. The OP parameter:
    This parameter is used by features like *BOUNDARY, *CLOAD, *DLOAD and so on. Its default value is MOD which means that said features are propagated (remain active) or can be modified with new values in subsequent steps. The non-default value NEW means that all features of a given type are removed unless they are respecified.
    It’s also important to mention that when boundary conditions are removed, the solver replaces them with equivalent forces and ramps them down to zero during the whole step (or removes them immediately in dynamic analyses due to a different default amplitude). When loads are removed, they are also ramped down linearly or removed immediately depending on the amplitude. Of course, the same applies to new prescribed conditions - they are ramped up or added immediately according to the default amplitude.

  2. Prescribed values:
    Magnitudes of prescribed conditions (BCs or loads) should always be given in terms of a total value, not as a change in magnitude. So BC set to 0.5 mm in the first step and 1 mm in the second step means that the nodes will move further 0.5 mm (to reach a total displacement of 1 mm) in the second step. If a magnitude is left unchanged (0.5 mm here) then nothing changes. Same with loads - they remain constant if there’s no amplitude.
    What’s important, setting BC magnitude to 0 in a subsequent step returns the nodes to their original positions.

  3. *BOUNDARY, FIXED:
    This parameter makes sense only in steps other than the first one. It means that, during the step, the specified DOFs remain fixed (frozen) at their values from the end of the previous step. Thus, any new magnitudes are ignored.

  4. Material state:
    Apart from the above rules, there’s one additional important aspect to consider. The state of the model at the end of a step is carried over to the next step. However, if the system is conservative (fully elastic) then there will be no impact on the results of that subsequent step. So each step is pretty much independent (like load case analyses if OP=NEW for loads) because there’s no permanent (plastic) deformation. The same applies to contact - if there’s no friction then the system is conservative so the solution does not depend on the load history but only on the final load.

4 Likes