Integrating FE and Dynamics

[Trying2Learn]

TL;DR Summary

Is it possible to use both codes (FE and Dynamics), independently?

(CAVEAT: I am on the verge of retiring. And as I look back on my life, I realize how much I do not know. And I am using the convenience of time, now, to go back and ask the questions I always wanted to understand. So, please forgive me for these questions. They have always been my mind and I am now trying to "really" understand.)

Suppose we have a robotic arm and we wish to calculate the stresses in the arm as the motor actuators move the various links.

Assume, initially, the arms are rigid. In such a case, we can turn to multi-body dynamics methods and calculate positions/orientations based on actuator forces/torques. This is fine.

Now, let me focus on one of the arms. Suppose I wish to know the stress in that arm (assume we are not modeling the arm as a thin member, but one wherein we actually can use the full 3D FE analysis -- i.e.: not a beam or a truss member, but a full 3D member; just assume that for the sake of my question).

We turn on the motor and do a time stepping analysis for the position of the arms, using forward (not inverse) dynamics/kinematics (say, Runge-Kutta or some such). We find the position/orientation of the arms as time progresses.

Immediately after we get the result for a specific time (say: tn , we turn to an FE code to conduct a stress analysis of the arm at that same time.

And we switch from one to the other, back to the dynamics, then FE, then dynamics, then FE and so on.
At tn, find positions and orientations from a multi-body dynamics code
At tn, find the stress from an FE code
At tn+1, return to the dynamics code and find positions/orientations
At tn+1, return to the FE code for stresses.
At tn+2 ...

Is this even possible?

In other words, if we turn to, say, Fung...

https://books.google.no/books?id=hmyiIiiv4FUC&pg=PA204&lpg=PA204&dq=how+does+fung+get+22+unknowns&source=bl&ots=OWfuV7sjyg&sig=ACfU3U2KuSt4J8BnP3_H2fcHSZsPKdtLvQ&hl=no&sa=X&ved=2ahUKEwiPv9epx_PmAhVkzqYKHcufCdQQ6AEwAnoECAYQAQ#v=onepage&q=how does fung get 22 unknowns&f=false

... we see the series of basic equations for linear elasticity, and they are highly coupled. But suppose we are not really interested in wave propagation of stress through the arm. Is it possible, for a quasi-static case, to do as I described above? What are the limits of doing what I described?

In which scenarios will my description above NOT work properly? What is lost by switching back and forth between two different codes (finite element and multi-body dynamics) (at each time step), to analyze the stresses or strains in a multi-body problem?

In other words, for stress wave analyses in, say, single bodies, I understand we must use dynamic FE analyses; but for a multi-link system, can we alternate the analysis between "rigid body" dynamics and then FE analysis? I am aware of emerging work in "flexible" multi-body dynamics -- beyond my ability -- but for some cases, will this approach work? Referencing back to the equations in, say Fung (above), are there any nonlinear, yet significant, terms that would be ignored by this approach?

 

[FEAnalyst]

To analyze a problem like that I would use one of the following approaches:
- flexible MBD in software like Simpack
- cosimulation between MBD and FEA software where positions and loads from rigid MBD are used in FEA, data is exchanged all the time (for example Abaqus with Simpack)
- dynamic FEA only - modern FEA software is no way limited to static analysis. It can perform advanced dynamic simulations with both rigid and flexible components. So it's possible to assume that some arms are rigid while others are flexible and simulate them in a single analysis. Here's an example:

Generally you can find some examples of all approaches if you look for engine piston or wind turbine cases. These are the most common.

 

[Trying2Learn]

To analyze a problem like that I would use one of the following approaches:
- flexible MBD in software like Simpack
- cosimulation between MBD and FEA software where positions and loads from rigid MBD are used in FEA, data is exchanged all the time (for example Abaqus with Simpack)
- dynamic FEA only - modern FEA software is no way limited to static analysis. It can perform advanced dynamic simulations with both rigid and flexible components. So it's possible to assume that some arms are rigid while others are flexible and simulate them in a single analysis. Here's an example:

Generally you can find some examples of all approaches if you look for engine piston or wind turbine cases. These are the most common.

May I follow up?

Thank you for your comments... I will look at the solution method you suggested.

However, I am not facing such a problem. My question concerns HOW one COULD solve it.

And for that, I must return to my original question. How are they solved?

Does the software you suggest solve a complete problem, time step, by time step, in which all aspects are advanced by a dt?

Or do they switch from FE to Dynamics, one after the other?

And, if the latter is the case, what is lost?

For example, I would imagine, say in the case of the robot arm, that if we do ONE time step with dynamics, then we do one with FE. But for the FE, there is a stress wave that passes along the link. That stress wave will have a role in the NEXT time step of the dynamics.

So HOW do they solve such problems? And by this, I do NOT mean: tell me exactly. Rather, even in your example of the cam pistons, do they alternate between codes (or, more exactly, methods)? Or do they solve some sort of equation system that THEORETICALLY integrates the deformation of multi-links?