I've imported a solid model of a parallel robot in abaqus and defined all the constraints, contacts, trajectories in joints and etc. It is moving reasonably. It's end effector has a little deviation from prescribe path. But the stress in some parts during simulation goes very high which seems wrong. I'm using dynamic explicit with automatic increment size in step. A point load has been defined on top of upper platform and tied to it. There is gravity too.
A video of my analysis has been attached.
I can't construct such robot and test it now! So how can I verify my results? Is there any similar work? Are flexible multi body dynamics in abaqus have been verified up to now?
Nice job indeed!
I am not expert in robotic; but from computational point of view, and as you mentioned correctly, you really need a validation for your model. Although you can rely on the already-verified-commercial FE codes (like ABAQUS) which mathematically can ensure you, you still need to validate or verify your model physically!
Personally I would recommend an analytical solution for your validation as it seems to be not that complicated to solve and derive the equations, replacing the deformable parts with spring (and dashpot) and rigid bodies with single mass! in that case you would be able to calculate the exact (let's say!) reaction force or kinematics!
And don't forget to not implement mass-scaling in your simulation to simplify the analytical simulation.
look forward to your validated model!
Apart from an experimental verification, which is the obvious immediate answer, it might be worth a try to reduce the time step - this might rule out numerical "flutter". Are there any numerical damping procedures running in the background of the simulation?
The mass propertied are not indicated, but stresses should indicate a proportionality to the externally applied load. Stresses in the actuators indicate bending; is this due to inertial effects, gravity or friction in the actuator end joints, or could the end joints be overly constrained?
Finally, I would suggest enforcing coincident rotational centers of the actuators (pairwise tops and bottoms) to make sure that the platform does not deform in bending.
The parallel mechanisms you are looking at is known as Stewart platform, which has been studied deeply in the past, at last for the kinematic problems (forward and inverse position problems, and velocity problems). Since it has been studied earlier, I think you can find the dynamic or kinetic studies. I know that must of the analysis will treat each body as rigid solid, so no deformation will be considered but at least can be give you an insight of what to expect from your simulation.
Are those prismatic joints? if so, why are calling it flexible, I can only see displacement from the video, and no actual flexion or twist, or any other type deformation....
But yes I agree with both of the replies earlier, you need to validate your model, and if it is validated analytically it will have a powerful conclusion!
Amin
Use Dynamic implicit and try to change the mesh (type, order and size) in those parts in which the stress is very high.
Regards,
End effector being deviated from the prescribed path could be due to the following reasons: 1 Joint co-ordinates 2. Damping and 3. Stiffness of links
High stresses could be due to fouling of links or parts of it...sharp edges of parts...arresting rotational dofs of the platform...
Selection of implicit or explicit integration scheme depends on the highest natural frequency, frequency range of interest, system is linear or nonlinear...
http://www.robotics-society.ro/noutati/2011.03-RA+%C3%9C%20Nadia.pdf
http://www.standardsmedia.com/Parallel-Robots---Mechanics-and-Control-7188-book.html
Thank you all
Some explanations to clarify may be useful:
Cylindrical joints have been defined under the moving pistons, so the forces go through the pistons and upper platform. The elastic deflections exist but we can not see them by eye!
I don't think there is any undesired arresting rotational DoFs.
No friction has been define in the model yet.
- Badges
- Science topic