RBE3 Element in NASTRAN 1

[kilbchoi]

Hi !

I have a question of rigid element usage of applying load.

Let's suppose I put the rigid cylinder weight onto the wide cantilever beam end, and I applied the side load at the CG causing the torsion on the beam. I have four or eight bolts to fasten the cylinder.

Most of people might use a rigid element, which has the independent grid at the CG and has four or eight dependent grids on the bolt locations, to model the cylinder.

What rigid element is most recommended ?

I experienced the very high additional in-plance reaction on the bolt locations is generated, if using RBE2 element.
I guess it doesn't allow the relative displacements of the bolt locations so that the high force is generated.

If I use RBE3 element, the additional reaction disappeared, since it allows the relative displacement.

I guess the latter is more practical since the former is too conservative, even if it depends on the stiffness of mating parts.

Anybody who can advise on this topic ?

Look forward to seeing it.

Thanks,

Kilbchoi

 

[pjhype]

If you really intended to model a rigid cylinder, it is appropriate to use the RBE2 element. Without knowing how you defined the weighting functions on each degree of freedom on the RBE3 element it is not possible to comment on the appropriateness of your modeling technique. Can you provide more details?

If you find that the RBE2 element is too conservative, I would recommend that you develop a more detailed finite element model of the cylinder. Using the RBE3 without good knowledge of the stiffness of the cylinder may introduce an unacceptable degree of uncertainty into your model.

pj

 

[kilbchoi]

Hi pj,

Thank you for your quick and great answer.

The reason why I think RBE2 is too conservative is,

(Input Model)
(1) Let's suppose, "y" axis is beam axis, "x" axis is perpendicular to "y" in the same plane.
(2) Beam length is relatively short, 300 mm length and 200 mm width, and it is made of 7075T6.
(3) My concerned load, 8g, is applied 230 mm below the beam plane in "x" axis.
(4) The cylinder body, actually not cylinder but sphere, is FLIR ball. (I am not sure you are familiar with it, it is special camaera attached to military or police helicopter)
(5) As I mentioned before, I have 8 bolts on the beam to hold the baby, approximately 50 kg. It is modeled by either RBE2 or RBE3.
(6) My beam designed by my great designer is "C" channel whose open section face up. As you see, it is torque-sensitive structure. That's why the "x" axis load is concerned.
(7) One beam end, loading side, is closed, but the other end is open. The other end is assembled to the column by gusset and butt joint. (Warping consideration is most headache in the analytical solution)

(Output)

With RBE2, the MPC's forces on the bolt locations in the beam are very unpredictable, and it looks unrealistically high. Actually if taking it, one of my bolts fails. As you see, the reaction force imagination is not simple since it is involved with stiffness distribution of the beam itself and with stiffness difference of two mating elements, etc. If making an example to show unrelastic load, there is extremly high "y" constraint force which is self-balanced with different signs on both sides, inboard and outboard. I guess it comes from two rigid modeling of in-plane filed as I mentioned before.

As a reference, the bolt pattern and the bolt size is deisigned by the vendor and has been sucessfully operated in the word. If you calculate the bolt load by hand with enough conservatism, your bolt load is quite low as well.

I am afraid if this reply is too long and confusing.

I really appreciate you get back to me with your great advice.

Thanks,

kilbchoi

 

[pjhype]

I do not understand how you will generate high Y-axis constraint forces if you load the model in the X-axis. I would think that you will get a shear force in the bolts in the X-axis that will be equal (on first order) to the total force divided by the number of bolts. You will also get an axial force in the bolts apportioned in a manner to create a couple that is equal to the applied moment. You will need to calculate this knowing the diameter of your bolt circle and the bolt pattern together with the Z-axis offset of your concentrated mass.

Try modeling the bolts with BAR elements. Connect one side of each bar to the channel (along the bolt pattern), and on the other side of the bars, create a spider (or cone) with RBE2 elements that attach to the concentrated mass. If you do not get shear forces in the X-axis and axial forces in the Z-axis, there is some other problem with your model.

pj

 

[kilbchoi]

Hi pj,

I've already tried your suggestion just before you replied to me.

(1) It looks much more resonable than direct RBE2 application. However it still has quite high "Y" constraint force.

(2) In this particular torsion problem, I concluded that the additional "Y" and "X"(not simple shear divided by number of bolts) has to be generated depending on the rigidity of mating element, in my case mounting pad of FLIR camera, since the internal load distribution is not simple.

(3) Complexity of my torsion beam
First of all, there is strong warping phenomena due to open "C" channel cross section so that there is transver shear distribution both on the leading and trailing edges, in my beam they are two flange sections. Please refer to Bruhn's book, Classic of aerospace structure guy. It says the torsion is absorbed by almost two couple transverse shear force on the fixed end due to warping constraint effect. (Obvously if it is uniform beam like circular rod/tube, no transverse shear exist but pure torsional shear exist.)

Therefore the rigidity of mating emement, FLIR pad, prevent relative in-plane displacement between bolts, and it causes the additional constiant forces.

Your brief estimation of the bolt load can be obtained by RBE3 modeling which sets CG as reference grid and eight or four bolts as independent grids. In this case, the bolt grids are independent each other and are affected only by the beam rigidity itself. In other words, they are freely moved in the plane regardless of mating element rigidity. The result is directly matched to what we can get analytucally with simple assumtions.

(4) I would like to suggest you to solve very simple problem as follows for the verification, if you are interested.

(4-1) Model cantilever beam by circular tube .
(4-2) Apply the transverse shear force by RBE2 and RBE3 which has a center grid and has relevent grids on the circumference.
(4-3) Compare both results. Obviously no difference in the global solution, stress and displacement of the beam. But, there is a hugh difference in MPC load due to similar reason to what I stated above.

(5) I learned the application of RBE2 to model any kind of load strongly depends on case by case. If the case is very uniform field, it shouldn't be a problem at all since no major error exist. However for the very ununiform field like my torsion beam, it has to be cautious especially for fitting analysys, even if no major difference exist in the global analysis, beam stress/displacement.


Thank you very much for your advice anyway and please advise further of additional background of RBE2 v.s. RBE3 if you have experience and knowledge.

Thanks,

kilbchoi

 

[brad]

RBE2 is truly "rigid"--it is constrained such that there is no relative motion between the associated nodes.

RBE3 is a constraint, but not really rigid. The single node's motion is dictated by the combined motion of the other nodes. This is a more flexible constraint, hence "rigid" is a misnomer when applied to the RBE3.

pjhype--Good comments.
Brad

 

[vonlueke]

kilbchoi: (1) Only use RBE2 if your camera mounting pad is truly nearly rigid compared to your C beam. Otherwise you could induce artificially high stresses, as you observed.

(2) It appears you might be violating a fundamental rule for RBE3, as follows. Avoid using RBE3 elements in areas of interest or areas having high stress gradient. Highly stressed areas may need more than even a cubic deformation to match a highly distorted, deformed shape, whereas RBE3 uses linear interpolation, which may be unable to describe the actual deformed shape. RBE3 works well only in areas of more or less uniformly varying stress, or areas having minor stress variations or low stress gradient...and away from areas of interest (i.e., in areas where accuracy of local stress results is unimportant).

(3) Since you are highly interested in the mounting pad attachment area, perhaps consider modeling your camera mounting pad base plate, also, as initially pointed out by pjhype, to simulate its actual compliance (and bolt loading distribution) more accurately, to see if your RBE2 result is reasonably correct or a false negative margin (whereas linear RBE3 in this scenario perhaps might be slightly unconservative).

 

[kilbchoi]

Hi guys,

I am happy to get comments from the top ranking persons in this forum regarding my case.

Hi vonlueke,

I want to confirm one of your recommendations, the number (1).

<1> At first it is not artificial high load in the perpendicular axis to the loading axis, in my case, &quot;X&quot;. It is naturally quite high loads depedning on the structure characteristics of &quot;C&quot; torsion beam and the rigidity of the camera pad.
<2> RBE2 causes the secondary constraint forces and it is quite reasonable, but it has to be cautiously considered since RBE2 is to model the FLIR infinitely rigid, and it causes too high secondary load on the fitting.
<3> I artificially modeled the camera with relatively thick plate connected to the CG location with RBE2, and the 4 or 8 bolts with BAR element as pjhype recommended. Obviously it is most reasonable modeling, and then I got corrected constraint forces, still high but acceptable based on the strength value.
<4> What I would like to point out from my case study, the secondary and unexpected, by analytical insight, constraint forces have to be created in such a ununiform internal load filed. For instance in the other load case to cause only the bending and the transverse shear force on the &quot;C&quot; beam, the order of the seondaty force is negligible and it is quite well correlated with our hand calculation. Torsion problem with ununiform cross sectin is really unexpectable and difficult porblem !!!

AND,

For the RBE3 comments, it is not right modeling as you and everybody pointed out.

Thanks

kilbchoi