Sorry I didn’t read what you guys wrote at the top.
Not sure if this could work?
If you define a Cylindircal coordinate system on a Dcoup3D node, a priori one could distribute a force in tangential direction (Second component) across the set of nodes coupled with him.
By other hand, Dcoup 3D can distribute the force according to user custom nodal weights.
Could those nodal weights reflect the mass distribution on the model.?
I’m curious about that simulation. Angular velocity seems too constant to me.
I think there is an imposed rotation on the shaft by means of Rigid bodies on each side.
What information is transferred from one step to the other? Displacements and Stresses or also velocities and accelerations?
If that’s the case, one could prepare an initial dynamic step in search of those conditions.
Note: Appling a force to a ROT node has a curious oscillating effect. Not sure what it is.
@vicmw “not too hard” for you would probably be very easy for Guido.
More importantly, lacking angular acceleration is a HOLE in CalculiX.
That means CalculiX has never solved an angular acceleration problem correctly and not likely to without that fix.
Simulations like below can never be done accurately with CalculiX.
The video is beautiful but is useless for engineering.
Thanks for your attention.
@Disla The crank in the piston-crank system is given constant rotation. For each instant, the position, rotation, velocity, angular velocity, acceleration, angular acceleration, joint forces, joint torques are passed to FreeCAD FEM workbench. In FEM, the joint forces and torques are converted to surface forces. Then Calculix calculates the static stresses in the connecting rod. So, no inertial forces are used yet. Nice picture but not engineeringly correct.
Now, we want to include inertial forces (body forces) from centripetal acceleration and angular acceleration to the FEM so the results are engineeringly correct. But Calculix doesn’t handle angular acceleration.
I don’t think Dcoup3D is sufficient because angular acceleration cause body force = element mass * R * Alpha which is a function of R distance from axis of rotation.
Sure it’s a hole. CCX is incredibly feature-rich so there aren’t a lot of holes. However, there are a lot of basic features that don’t work correctly or have surprisingly severe limitations (see most of the issues on Github). There are also holes for more commonly used features such as surface traction and RBE3 but I can understand that since you’re making software that needs general linear + centripetal + angular acceleration, it must be frustrating that one of them is missing.
I’ve also been frustrated by missing features and bugs so I maintain my own fork and just do them myself. It’s much more effective than banging your head against a wall hoping for someone else to when they have different priorities.
It’s a stretch to say CCX has never solved an angular acceleration problem correctly. It can still do it with dynamics.
Dynamics! That may be what I need. I just need to solve for the stresses at the initial conditions given position, rotation, velocity, angular velocity, joint forces, joint torques and gravity. Is there an example? Thanks.
I checked DYNAMICS and STEADY STATE DYNAMICS. They are for small displacement in an inertial frame. So, no rotating frame. Definitely, Calculix have never solved angular acceleration.
Your approach is certainly new to me. Seems you are transferring variable states from MBDyn to the linear solver to get the stresses. Why not solving directly with nonlinear dynamics in ccx?
I guess your path it’s computationally more efficient.
I gave an oportunity to ccx to see what it does with rotating system and Fictious forces seem to emerge naturally.
NOTE: I haven’t apply gravity as I don’t see it’s role in this problem.
Thanks. Now make the part very stiff almost like a rigid body and see if Calculix can solve it. It would be almost impossible.
Perhaps STEADY STATE DYNAMICS might work for stiff bodies. Can you try that? Thanks.
It solves even faster. Thats normal. It’s almost a Rigid Body.
I don’t think this problem can be solved with Steady State Dynamics. I’m accelerating the system. There is no steady solution.
I have set up to reach close to 1500 rpm in 1 second which is a lot. (Gif displays 1 second in 20 seconds to be able to see something)
It could be similar to the effect of the transient startup with a variator.
The first model was Steel material parameters.
This one is made much stiffer.
EDIT: Computation time is 1min 45seconds with 192 Elements / 413 nodes as reference.
Can you see stress variations in the stiff body?
You are setting up the problem correctly.
Can you make the color scale to show the stress variations.
I think STEADY STATE DYNAMICS mean that the internal vibrations have died down.
So rigid body accelerations are ok to use. This is what I want if the stresses are calculated.
Very fast Stress variations due to excited high modes + coarse mesh + deficient time step.
-I have look at the main oscillating mode and cut it with Raileigh Damping proportional to inertia.
-I have change to C3D20R and refine x2.
-I have set up Steel material properties and more reasonable ramp up to 400 rpm.
See vid after damping.
There is a mixture of tension and bending stresses present in the model.
First, bending dominates. Looks like the cantilever beam Stress Distribution.
As omega increases centrifugal force starts to show up unbalancing the compressive/tensile distribution.
-I think STEADY STATE DYNAMICS mean that the internal vibrations have died down.
If I’m not wrong Steady State solution is build guessing the solution can be built as a superposition of a discrete number of modes. Internal vibrations (high modes) disappear because one use a limited number of modes, let’s say 30 (Lower ones). It acts like a filter. By other hand , you are asking to impose an angular acceleration ¿isn’t it?. ¿How could a Steady State solution be representative of a Transient phenomena?. With constant rpm maybe but with angular acceleration I don’t see its posible. What do you think?
Great progress. To check the answers are correct, let’s do a very simple example.
Make a horizontal beam: length(x) 1000mm, width(y) 20mm, height(z) 10mm
Material: uniform steel
Applied forces: 1000N in z direction at 0,0,0 face. 1000N in z direction at 1000,0,0 face.
No displacement constraints.
No gravity.
Expected answer: Beam stresses should behave like beam pinned at ends and under uniform gravity.
Do DYNAMIC and STEADY STATE DYNAMIC.
DYNAMICS should show transient vibrations superimposed on beam displacement under uniform gravity like load.
STEADY STATE DYNAMIC should show only beam displacement under uniform gravity like load.
Thanks.
I want to learn CalculiX too. Can I see how you simulate the problem? Thanks.
I don’t understand your numbers.
Your bar weights 1.56 Kg and you are applying 1000N on each side ?¿?
Regarding the expected response, Dynamics do not necessarily should show transient vibrations. Recall Impulse and how are you applying the force.
FeAnlyst has some nice tutorials about Dynamics and Modal using Prepomax and Calculix.
You can follow them and inspect the inp.
They can help you to fullfill your goals.
Yes. The beam will be accelerating. Hence it will have a body force that feels like gravity.
DYNAMICS should be able to do that. Correct?
Your bar is under like …130g ?¿?. Static and Dynamic can’t be comparable. Dynamic analysis will probably fail due to large deformations or will require tiny timesteps.
By other hand, Pined support doesn’t have the same BC as pulling up. I have consider a roller at one side.
g=10 m/s2
density=7850 Kg/m3
Adjusting that ,I solved it and it gives excellent agreement in between the Stresses of the three different methods. No internal modes oscillating around.
Here it’s a rectilinear acceleration. Regarding the same problem but involving rotations, I don’t see it. Please someone correct me if I’m wrong. Dynamic systems is not my area.
Static
Modal Dynamic. 50 Modes. Ramped load.
Does “roller at one side” mean roller moving in Z direction only and rollers are pinned to the beam end at X = 0? You also have 1000N force in Z direction at both ends?
What are the “three different methods”? What midpoint deflections did you get for each?
Thanks.
Roller one side, Pinned at the other
No. 7.85 N to compare with the beam under self weight.
I’m using C3D8R.
- Nonlinear quasiestatic, 2) Dynamic (Implicit) and 3) Modal Dynamic
I’m using the fastest C3D8R. That should be inspected. Was just to speed up the test.
Results | |||
---|---|---|---|
1) Nonlinear quasi static | 0.564 mm | ||
2) Dynamic (Implicit) | 0.57 mm | 1636.09 mm | 1635.52 mm |
3) Modal Dynamic | 0.57 mm | 1667.03 mm | 1666.46 mm |
How can I send you a *.pmx file to try? The forum forbid uploading *.pmx. Thanks.