Dynamic Implicit vs Static General

[ToTheMoon]

Can someone help me to understand why in certain non-linear models the problem can be converged more easily using a dynamic, implicit quasi-static procedure rather than static, general?

 

[FEA way]

In dynamic implicit quasi-static step inertia effects are introduced to help with instabilities and those are one of the most common reasons of non-convergence.

 

[ToTheMoon]

When you say the inertia effects are introduced, under quasi-static are they primarily introduced just to stabilize? My understanding is if you have a problem that you can converge in static, general and run the same problem as dynamic, implicit quasi-static you will get the same results.
Or is it alternatively that if you are using this approach you already know the inertia effects are negligible for the overall solution so applying them in the dynamic, implicit step does not overly affect the solution?

 

[FEA way]

Yes, inertia effects in implicit quasi-static analyses are used mainly for stabilization. We often choose this type of step when we are looking for a static solution of an unstable problem (such as buckling) that causes issues with convergence in general static simulation. For stable problems pure static analysis is usually more efficient.

 

[ToTheMoon]

Perfect thank you! In terms of Dynamic Implicit with and without quasi static applied is the main difference how the forces or displacements are applied within the step as well? So for instance with quasi-static if you have a load applied as a ramp function where the ramp time is equal to the step time the load will be applied slowly over the step time to simulate static loading. Whereas for the same case without quasi-static applied the load would be applied quickly and you would see oscillations in your system?

 

[FEA way]

The differences between each dynamic implicit step application are more subtle and relate to the time integration method being used:
- transient fidelity: Hilber-Hughes-Taylor time integrator with slight numerical damping
- moderate dissipation: Hilber-Hughes-Taylor time integrator with moderate numerical damping
- quasi-static: backward Euler time integrator

This is described in the documentation chapter "Implicit dynamic analysis using direct integration".

 

[ToTheMoon]

Great thank you, I'll read up on that section. If you do not specify an application type does it default to transient fidelity?

 

[FEA way]

It depends on whether or not there is contact in the model. Analyses with contact are set to moderate dissipation while those without contact are set to transient fidelity by default.

 

[ToTheMoon]

Much appreciated thank you!