structural analysis with patran-nastran, warping display

[MichMik]

Good evening to all

I am a new user of the forum, I opened this discussion because I have a problem with the analysis of a structural element in patran and I don't know how to solve it, I hope you can and want to help me.
in practice I have to analyze a structure that has the form of a rectangled rectangled parallelepiped subject to a torque moment at one end and totally stuck at the other end; therefore it is a beam with rectangular hollow section with thin walls. I shaped it with quad4 elements with shell properties. the problem arises when I have to apply the torque moment, I tried to create a knot at the center of the rectangle and to connect it with all those of the final section of the beam with rbe2 elements and then I applied the torque moment to the central node, but I read somewhere that the rbe2 elements introduce an additional rigidity in the section and maybe this is the reason why, when I analyze the results, the war section instead I also thought of using rbe3 elements, but I don't know exactly how to get with them a simulation of the structure's behavior as close to reality as possible.
Thank you for your availability.

 

[Onda]

You can apply forces, equal and contrary, applied to the sides of the rectangle, taken at 2 to 2, so that the sum of forces balances, but remains the torque moment.
or, you can use a rbe3, whose central node, dependent, must be connected from 1 to 6, while the independent beam boundary knots must be connected from 1 to 3.

otherwise, third solution, attach your beam of a length enough to be able to overlook the additional stiffness and use a rbe2., the torque moment is constant and the length of the beam does not change the result.

 

[MichMik]

I thank you for the attention, I think I will adopt the first proposed solution. If I may ask why in case I use rbe3 elements, I should bind the central knot from 1 to 6 while the independent knots from 1 to 3. understand it would help me to use this element in the right way even in the future.
Thanks again for the availability.

 

[Onda]

exact, that is the mode of use of rbe3, as a translation of the independent node, turns into a rotation of the dependent node.

 

[MichMik]

If I used rbe3 by binding only the rotations of both the dependent and the independent ones, would I get correct results?
I wonder why I just tried to use rbe3 in the proposed way, but I get a strange deformed of the final section of the beam. the central part rotates, while the short sides of the rectangle tend to remain in their position (translate only slightly), so I find a sinusoidal waveform of the rectangular section in its plane. I wonder if the reason does not lie in the fact that the rbe3 elements transmit the load differently to the short sides of the rectangle, which are more distant from the center, than the load transmitted to the central part of the section and to the long sides.
Thanks again for everything.

 

[Onda]

no, you don't have to use rbe3 by turning in independent nodes as they are transferred strangely to the dependent node. rbe 3 are a little peculiar and it is not said that they distribute the load exactly as desired.

 

[MichMik]

Finally I chose to resort to the first solution, that is to break the torsion load in two equal and opposite forces distributed on the short sides of the rectangular hollow section. the deformed that I get seems likely to me and in fact the rectangle cable not only rotates but is deformed by warping. I will compare this solution with that obtained by adopting the second method, i.e. using the rbe3 elements and binding the dependent knot from 1 to 6 and those independent from 1 to 3. at first the deformed obtained using the rbe3 elements seemed unreal, but perhaps not so much, evidently warping when it is free to manifest deform the section in such a strange way that it does not seem to have been applied only a simple torguing moment.
Thanks again for your help.

 

[Onda]

Finally I chose to resort to the first solution, that is to break the torsion load in two equal and opposite forces distributed on the short sides of the rectangular hollow section. the deformed that I get seems likely to me and in fact the rectangle cable not only rotates but is deformed by warping. I will compare this solution with that obtained by adopting the second method, i.e. using the rbe3 elements and binding the dependent knot from 1 to 6 and those independent from 1 to 3. at first the deformed obtained using the rbe3 elements seemed unreal, but perhaps not so much, evidently warping when it is free to manifest deform the section in such a strange way that it does not seem to have been applied only a simple torguing moment.
Thanks again for your help.

I think it's more unreal as you simulated the forces, putting them only on the short sides. I would have put them on all four sides, of force proportional to the length of the side, so that their sum gives the time required. Whereas in a thin wall section the cutting flow due to the twist is constant. .

 

[MichMik]

I had thought of breaking the torque moment into four forces distributed on the four sides, but then I remembered that acting in this way I would have imposed a constant cutting flow in the final section of the beam and I did not want to do so. in fact only if the structure does not have constraints that prevent warping the cutting flow is constant. In my case, the disaster limits warping and cutting flow in all sections, so also in the final section, is altered compared to that provided by classical theory.

 

[Onda]

Then I don't understand anything... If you have a bond, put a rbe2. and the section remains flat, otherwise you do not have a constraint, and the section is free to deform. maybe if you put an idea of what you want to schematize, then you see the best method of representing it

 

[MichMik]

I'm sorry, I didn't explain.
In practice I have to analyze a rectangular hollow beam with thin walls, bound to one end and free but with a torque moment applied to the other end. It prevents warping in the initial section and alters cutting flows in all sections, which should take values other than those foreseen by classical theory.
I thought that the mathematical torque moment is an integral bond, being a characteristic of the stress, that is, I should impose that in the final section of the beam the integral of the cutting flows is equal to a known value (that of the torque moment), but the cutting flows should let them free to assume in the section any value provided their integral is fixed. I don't think there's a way to impose such a bond on Patran.

 

[Onda]

I keep not understanding your explanation, "victed at one end and free but..."
Then I don't understand what you mean by imposing such a bond.
a bond is a value (0 or different, no matter) to a degree of freedom of the knot. if you don't want to impose all 0, calculate nodal shifts, and imponile.
But I think we're getting farther away. .
put a sketch of what you have to do, because it is getting darker

 

[MichMik]

I meant that the rectangular hollow and thin-walled beam is bound at one end with an ink and not bound at the other end, where only a torque moment is applied.
the torque moment is a characteristic of stress, so it is a generic external load, in the sense that it could be generated in various ways, because there are endless distributions of cutting flows that integrated produce the same moment. the constant cutting flow predicted by classical theory is one of these distributions. I should impose in the final section that the distribution of cutting flows is such as to produce the torque moment to which I want to subject the structure, but before the analysis I should not impose other conditions on the cutting flow, which must be a result to obtain.
If I break the moment in two or four equal and contrary forces distributed on the sides of the rectangle, it is as if at that moment I was blocking in the final section the trend of the distribution of cutting flows to the value established by me.
the attached file is a pattern of the structure that I have to analyze.

 

[Onda]

mah, the fem is made to analyze real structures, not unreal structures, you somehow torque you have to give it, and this somehow affects the tensions around the load point, this is a reality.
Then, given the length you will probably have an area in the center where your torque follows perfectly de saint venant, as you are far enough from the constraints and areas of application of the load.

 

[MichMik]

Yes, in fact I am realizing that with the fem it is not possible to impose a characteristic of stress in a generic way, as it is used to do mathematically when studying the same structure with analytical methods. It is necessary to specify how the external load is distributed on the nodes of the section on which it is applied. when explicit how the load is distributed on the knots inevitably a choice is made.
Thank you for all the attention you have addressed.