cran, f06, epsilon too big

[drpulsae]

nice users of the forum. I have another question to propose. from a few months model with patran. the last two structures I modeled (aluminum frames) have a big problem: reading f06 about the load case of the analyses I launched I found an epsilon too high. in one is of the order of 10e-3 (at the limit of the acceptable), in the other even of 10e+3!! ! ! !
This result, especially the last one, is shocking. I can only tell you that the utility verification does not detect any problem sensitive to the elements (the model consists of elements 1d bar and 2d shell quad4 and tria3 for a total of cbar: 1387; cquad4: 122245; ctria3 7448; cry: 167998; rbe2: 683) (no element fails the test aspect ratio, taper etc...). welds were modeled making an equivalence with the firmness of combining the knots of the interested parties and bolts were modeled by means of mpc rbe2 connected by bar elements to the master nodes. constraints were applied to the master nodes of some mpc. loads are inertial. I tried to create a dof list by selecting all the degrees of freedom and knots of the whole model but it remains empty. the analysis is however good and shows results, unfortunately not significant due to the high epsilon values. I have three questions:

1) How can I control the epsilon value?
2) which tool do you recommend me for the quality assesment of the fem model?
3) using this tool, can I trace back to the origin of the problem?

are available to provide all the details that can help you find an answer. thanking you in advance, I look confident.

 

[Onda]

here you will find a series of controls to be carried out:http://femci.gsfc.nasa.gov/validitychecks/index.htmlWhat measuring unit did you use for your model?
if the loads are inertial, and the accelerations are in mm/s^2 the masses must be in tons and the density in tons/mm^3 otherwise the units are not congruent.
if you use the masses in kg you have to put the wtmass parameter 0.001 in the bdf to scale the masses and density
otherwise your loads are multiplied by a thousand factor.

another reason that could raise a lot of epsilon (I think you mean strains with epsilon) is if you have thermal loads and over bound the model.
you may also have mistakenly grounded knots and created hyperstatics, with a restrained modal as described in the link I poostato find the first six rigid ways and verify the model

Then, if you put some picture of the model, and some extra indication.

greetings

wave

 

[drpulsae]

wave, you're kind and solert. This is the cursed string of the f06:

_
for data block kll
load seq. no. epsilon external work epsilons greater than 0.001 are flagged with asterisks
1 2.5522400+03 8.1765121e+01

the units of measurement of the model are mm, ton, mm/s^2, congruent. by epsilon I mean the measurement of the numerical accuracy of the fem model. I read that its order of magnitude indicates the significant figures of the solution: since the model is in mm acceptable solutions should bring the order of the meter (absurd!). the mesh has the average side of 5 mm. I launched a modal analysis (only 103 subcases: default) and I got 10 ways of the structure, including between 11 and 35 hz. example of constraints:Saldature:there are no thermal loads. the constraints I have created are incastri: prevent translation and rotation and are applied to the nodes: 16 master mpc nodes connecting to the rest of the axieme and 4 which are knots end of bar elements.
this is the result of the modal analysis of the private model of constraints, as described in what you post:and this is a quick plot of a solution:

 

[Onda]

I think you have a problem,
effectively that epsilon value must be low, below the explanation of what is epsilon and why it has to be low.
try to delete the constraints and turn a modal, to see if you actually have 6 rigid ways (the first six frequencies must be close to 0). check that you do not have cbush or celas of length not anything that introduce fictitious loads.
verify the orientation of the beams.

View attachment 39871