The reference point is rotating, and I want to measure its vibration displacement during rotation. Perhaps because this model only has one gear, I will try another pair of gears.
I am working on a meshing simulation under constant speed and torque of a pair of gears. After several days of attempts, my simulation results are still unsatisfactory. As shown in the figure, the stress results of the gears are too high, far exceeding the stress results from ANSYS. I don’t know where the problem lies. I have tried modifying the contact parameters, adjusting the mesh density, modifying the minimum time increment step, and switching to implicit methods, but none of them have been effective. I hope you can provide me with some guidance. Attached below are my original files and the results of the simulation.
I’ve mentioned it here: How to apply rotational speed correctly - #5 by FEAnalyst
In explicit dynamics, you may easily go from quasi-static to fully dynamic response if the loading rate is too large or mass scaling is too high. Your gears clearly vibrate and it’s not a quasi-static analysis. You should request the output of kinetic energy. It should be a small fraction (< 5%) of total/strain energy.
I don’t quite understand, are you saying that my simulation is fine, but I just chose the wrong output object? Then why is the axial vibration displacement of the gear much larger than ANSYS? My ultimate simulation goal is to obtain the vibration displacement and mises stress.
It depends on what you want to achieve. Typically, quasi-static analyses of such systems are carried out and then the kinetic energy should be very small. You can perform a dynamic analysis too if you are interested in vibrations but you have to be very careful in the case of explicit dynamics. Currently, you force the gear to rotate by around 900 degrees in 0.01 s which means 250 rotations per second.
1 Like
In the current scenario I want to simulate, the gear conditions are as follows: the rotational speed is 250 revolutions per second, with a load of 2 Newton meters.
Ok, that’s way faster than I assumed. But then you basically shouldn’t use any mass scaling at all. And it would be good to compare the results with implicit dynamics if it converges.
So how should I set up this high-fidelity dynamic simulation now. Does large-scale scaling refer to adjusting the time increment steps
Mass scaling in CalculiX is triggered by specifying a non-default minimum time increment so you should leave the default setting for this parameter. It will probably slow down the analysis a lot but it’s pretty much unavoidable in this case with explicit dynamics.
-
As shown in the figure, after setting the minimum time increment to its default value, the simulated stress values remain significantly high, with little difference compared to yesterday’s results.
-
After changing the mesh from tetrahedral to hexahedral, with all other settings the same, the model with hexahedral mesh encountered an error, which is quite strange.
-
Additionally, I’d like to inquire about how to output the rotation speed.
Don’t use Direct incrementation here, leave it as Automatic and just don’t change the minimum time increment.
Perhaps the model became too large and a different solver (like Pardiso) could handle it.
Nodes of solid elements don’t have rotational DOFs so, like with displacements, only the translational velocity components are obtained. You can get rotational ones from reference points though.
I have adjusted the time increment by Settings in the last two days, but the simulation output stress is still very high, should I continue to use the dynamic module, Or replace the other modules?
Dynamic implicit can be a better choice but you have to keep in mind that there might be convergence issues in this procedure, requiring the adjustment of contact properties, usage of different BCs, small numerical damping or even soft springs.
Thank you for your reply. I will try your suggestion.