best way to make a distance only constraint between rbe elemts?
I'm trying to solve nonlinear contact analysis for Chain drive with 2 sprockets drivetrain system. This is what giving me a hard time now.
The way I modeled my chain is using shell component with rbe2 dependent node on center of chain shell and independent nodeson each nodes on shell. After I linked all the shells with several types of rbe's(rbe2,rbe3,rjoint) between the dependent nodes in chain shell. The result shows each chain shell does not constraint enough. I wonder is there anyway to give a distance only constraint between 2 rbe elements?
the chain model contacted to drive sprocket didn't give a type of effect on driven gear contatcted chain model.
I failed to converge the NL so I tried Linear Static first and got a problem.
the sprocket looks like it is just not rotating and the size is keep increasing. I can't find out what's wrong on my model. Is there any possible cause for this?
Thank you.
Best Answer
-
In the linear static model there is no 'rotation' (hence the 'linear'!) so this is expected, I replied to your other ticket (but I think you have deleted it) you may need to move to explicit simulation for this loadcase
1
Altair EmployeeAnswers
-
In the linear static model there is no 'rotation' (hence the 'linear'!) so this is expected, I replied to your other ticket (but I think you have deleted it) you may need to move to explicit simulation for this loadcase
1 -
how about non linear static?Paul Sharp_21301 said:In the linear static model there is no 'rotation' (hence the 'linear'!) so this is expected, I replied to your other ticket (but I think you have deleted it) you may need to move to explicit simulation for this loadcase
0 -
Yu JH said:
how about non linear static?
nlstat is only material and contact non-linearity (i.e. plasticity) not geometric non linearity, so the same issue would occur, lgdisp would be needed to account for rotation
1 -
Thank you i would try and update soonPaul Sharp_21301 said:nlstat is only material and contact non-linearity (i.e. plasticity) not geometric non linearity, so the same issue would occur, lgdisp would be needed to account for rotation
0 -
1 more question, if I put 10 in ninc in nlparm(lgdisp) will it be a nlgeom?Paul Sharp_21301 said:nlstat is only material and contact non-linearity (i.e. plasticity) not geometric non linearity, so the same issue would occur, lgdisp would be needed to account for rotation
0 -
Yu JH said:
1 more question, if I put 10 in ninc in nlparm(lgdisp) will it be a nlgeom?
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
1 -
Paul Sharp_21301 said:
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
I checked .out and did 1-d elements dependency check all of the rbe2 I used for chain links were double dependency so I changed it to cgap card and tried solve it and this came up.
*** INFORMATION # 3403
MUMPS failed in factorization because the run-time memory
requirement was larger than the estimated amount. Solver will
retry with a larger memory relaxation coefficient.
*** WARNING # 3400
Singularity detected at following 1 DOFs.
Possible reasons are:
1) insufficiently constrained model,
2) having rigid body mechanisms within the model,
3) extremely ill-conditioned rigid element sets,
4) extremely thin shells (as used for skinning) that have MID2/MID3,
5) gap elements with extremely high stiffness (KA, especially KT or MU).
6) extremely high Poisson's ratio in hyperelasticity material definition.
Check the model and rerun the problem.
(MECHCHECK may be used to find the rigid body modes. To do so,
change the input to be an eigenvalue analysis and add MECHCHECK.)
(WARNING: results obtained with MECHCHECK cannot be used because
the model is changed internally.)
------------------------------
Grid No. Component
------------------------------
121065 3
------------------------------I'm trying to find Grid no. but I guess there is no way to find it. And also I have put lgdisp with 10 and the model looks like this still the model just keep increasing it's size.
0 -
Paul Sharp_21301 said:
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
I will try springs right now.
Huge thanks for help me out.0 -
Paul Sharp_21301 said:
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
I have encounterd a problem I'm trying to use spring elements and It should implement chain links but the srping elements have option to check dof's. for the chain links it's a kind of hinge so should be rotatable.
0 -
Paul Sharp_21301 said:
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
I don't need that much rotation.
I'm trying to see the situation about the car initial accerlating so I put the point mass on the driven gear's rbe2 dependent node. I guess the rotate needs lower than 1 degree0 -
Paul Sharp_21301 said:
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
same things happening now I think I should review my whole model and check it
0 -
Paul Sharp_21301 said:
yes, well it will be lgdisp (nlgeom is another different, older and now deprecated solution sequence), it will try and solve over 10 increments/steps (but step size may change in run, so all it really sets is your first step initial try at 10% of total load)
I don't know how much rotation you are expecting? this may not be enough
Going back to your original question, it may be better to replace the rbe2 'between' links with springs?
Can you please check my model?
0 -
Yu JH said:
That still seems to be using the chained RBE2? If I have time over the weekend I'll take a look to see if I can set up in a way that works
1 -
Paul Sharp_21301 said:
That still seems to be using the chained RBE2? If I have time over the weekend I'll take a look to see if I can set up in a way that works
Thank you very much
0 -
Paul Sharp_21301 said:
That still seems to be using the chained RBE2? If I have time over the weekend I'll take a look to see if I can set up in a way that works
I've got an information about lgdisp-lgdisp_v1 as 1 on param card for all load step so I put that card and It's solving without error yet. I think it will take some times to solve with my old pc. So putting lgdisp on param can make a difference in this case?
0 -
Yu JH said:
I've got an information about lgdisp-lgdisp_v1 as 1 on param card for all load step so I put that card and It's solving without error yet. I think it will take some times to solve with my old pc. So putting lgdisp on param can make a difference in this case?
that shouldn't make any difference!
setting LGDISP above or in the subcase should be the same, maybe you have an older version of OS, changing that doesn't get the model you sent running for me, behaviour is the same. But if you have a working version now that is good
1 -
Paul Sharp_21301 said:
that shouldn't make any difference!
setting LGDISP above or in the subcase should be the same, maybe you have an older version of OS, changing that doesn't get the model you sent running for me, behaviour is the same. But if you have a working version now that is good
I have OS 2022.1 and probably this is a newest one as I know. I also tried OS explicit using the card NLEXPL, and the solver keep calculating while cycle goes just keep increasing forever. Also the output file has only showing 2D graph. So I dont know what's going on inside my model. I'm gonna learn about NLEXPL and update soon.
0 -
Paul Sharp_21301 said:
that shouldn't make any difference!
setting LGDISP above or in the subcase should be the same, maybe you have an older version of OS, changing that doesn't get the model you sent running for me, behaviour is the same. But if you have a working version now that is good
I got rid of EXPERTNL on PARAM card and changed BC's to SPC's on every chain link. I've put NLEXPL and give the time TTERM for 0.0001 sec. It ran without problem and also took just few minutes!! I guess EXPERTNL was the main problem. I've changed model to simplify and for the successful run so I need to go back to the previous one and see what's going on.
0 -
Yu JH said:
I got rid of EXPERTNL on PARAM card and changed BC's to SPC's on every chain link. I've put NLEXPL and give the time TTERM for 0.0001 sec. It ran without problem and also took just few minutes!! I guess EXPERTNL was the main problem. I've changed model to simplify and for the successful run so I need to go back to the previous one and see what's going on.
This is my best effort at an NLEXPL run
A few things, your point mass at the centre of the large sprocket had no inertia values, so the whole model had no additional resistance to rotation apart from the inertia of the sprockets and chain themselves I added some nominal values. for inertia to the point mass, but they are not calculated from anything.
Because the chain links were penetrating the sprocket I set clearance 0 in a PCONT
Chain 1d part I swapped out for CBAR/PBAR links with steel props
Then because pinflags are not supported for CBAR in explicit, each 1d link is joined to the next and to the pins with cbushes
All the values used for chain links and bushes are arbitrary, but you can use it as a guide, rotation happens!, run takes a couple of hours.
Whether this is a valid route is debatable, perhaps Radioss would be a better choice for this analysis.
1 -
Paul Sharp_21301 said:
This is my best effort at an NLEXPL run
A few things, your point mass at the centre of the large sprocket had no inertia values, so the whole model had no additional resistance to rotation apart from the inertia of the sprockets and chain themselves I added some nominal values. for inertia to the point mass, but they are not calculated from anything.
Because the chain links were penetrating the sprocket I set clearance 0 in a PCONT
Chain 1d part I swapped out for CBAR/PBAR links with steel props
Then because pinflags are not supported for CBAR in explicit, each 1d link is joined to the next and to the pins with cbushes
All the values used for chain links and bushes are arbitrary, but you can use it as a guide, rotation happens!, run takes a couple of hours.
Whether this is a valid route is debatable, perhaps Radioss would be a better choice for this analysis.
Great thanks to your support! I was having hard time cause there in no such a guide line about NLEXPL on any websites. This one would be really useful.
0 -
Yu JH said:
Great thanks to your support! I was having hard time cause there in no such a guide line about NLEXPL on any websites. This one would be really useful.
Yes, NLEXPL is new (only really available for last version or 2) and constantly developing, there is an overview in help: NLEXPL in help
I expect there will be more examples once the development has plateaued a bit!
1