Nastran - Delumping of Internal Loads 1

[VN1981]

Hi,
It seems like we (employer) might have missed an important step in extracting internal loads from Aircraft FEM models...delumping! I found Ian Tiag's fantastic Aircraft FEM notes and I came across the concept of delumping of internal loads when 2D-1D elements are used for web & chord/stiffeners representation. I've spent the last 2 weeks going over the same material & I was able to dig out more information about the above procedure, but I still have questions. Some of them may sound obvious & stupid...so please bear with me

1. Is delumping procedure in use even today in major OEM practices?

2. Is delumping performed for any combination of 2D-1D (Cshear/Plate-CROD/CBEAM) or is it particularly reserved for CROD element I am assuming the answer is YES, but just looking for clarification from experienced users.

3. If I have lets say 30 elements (30 each of 2D representing SKIN & 30 each of 1D representing STR) along the span of wing between 2 rib stations and if I have to perform delumping manually, it becomes tedious. Perhaps the big OEMs have in-house tools or Excel macros to extract/communicate information from NASTRAN and spit out delumped loads, but any tips on how to perform delumping operation with minimal effort?

4. Lastly, any open source material where I can learn more on the above procedure?

Thanks in advance,
- VN

 

[rb1957]

I think "delumping" is applicable when you model effective skin as part of the stringer area (and have CSHEAR panel elements). I would have thought this to be a very "olde" way of modelling.; that today we model fully effective skins and hand calc the compression loads (modify the compression stresses to account for effective skin).

one place I work at deliberately "lumps" the FE load, they combine stringer loads with the compatible skin load at each node (in preference to using the native FEA results). Then they carefully "delump" the load via hand calc.

another day in paradise, or is paradise one day closer ?

 

[VN1981]

Thanks for responding.

I have a few follow up questions.

Let me start with concept of effective skins. Per my understanding (from Bruhn, Flabel & OEM Manuals), the concept of effective skins kicks in only during post-buckling of thin sheet panels. Lets take a wing and if I've understood the normal practice, at limit load there should be no gross permanent deformation which can impede safe operation of the structure. My understanding is that, at limit load the wing skin panels (& also stiffeners) should be capable of resisting compressive loads without buckling i.e. they are fully effective. Post limit load, once the skin buckles, the concept of effective skin kicks in for calculation of crippling and/or buckling allowable for Skin-STR/stiffener combination.

So my question is, where does effective skin play a role in linear static analysis? I want some clear understanding. I haven't done a post-buckling analysis yet, so can't speak of how FEM results are used in sizing.

>> that today we model fully effective skins and hand calc the compression loads (modify the compression stresses to account for effective skin).

Can you please elaborate if possible with reference to what I've posted about effective skins?

I checked with an ex-colleague of mine who works at a major OEM who are in to Thrust Reversers and she hadn't heard about delumping. So, I guess it depends on the company practice & probably isn't in vogue today.

 

[rb1957]

1) what is "effective skin" in compression ? (I know, 30t) The point is that a FEM, with fully effective skins, will underestimate the effective stress in compression (because it over estimates the effective area).

In the olde days they modelled with RODs and CSHEARs, and included the effective skin area in the ROD area. Hence the need to "delump" loads between the different members.

These days we'll model the stringer as itself and the skin as itself ... no need to "delump" (as I understand it) but now we have to be careful with compression stresses. We can take the full compression load from the model (like from GPFB), and hand calc it over the stringer and effective skin.



another day in paradise, or is paradise one day closer ?

 

[VN1981]

RB1957,
Thanks again for answering. Thanks for clarifying delumping of loads from Nastran output.

I think I may have used the term "delumping" incorrectly while asking the question. I apologize for the confusion if any. My original issue was with splitting the GPFB of 2D element (CSHEAR or PLATE/QUAD4) in to shear & end panel loads (ref. page 17/72 of Ian Tiag FEM notes pdf). And adding the shear term to the GPFB axial load terms of 1D rod/beam element and thus obtain a varying axial load in the stiffener/STR (Ref. page 9/72 Ian Tiag pdf). The procedure involves obtaining the GPFB at each node of the element & calculating shear contribution (1/2 x SUM of loads in each direction) etc...

I hope I've made my concern clear this time. With reference to the above, let me ask my questions again.

1. Is the above procedure of splitting into shear & end panel loads in use even today in major OEM practices?

2. If there are lets say 30 elements (30 each of 2D representing SKIN & 30 each of 1D representing STR) along the span of wing between 2 rib stations and if I have to perform delumping manually, it becomes tedious. Perhaps the big OEMs have in-house tools or Excel macros to extract/communicate information from NASTRAN and spit out delumped loads, but any tips on how to perform delumping operation with minimal effort?

3. If the answer to question #1 is YES, then we cannot use stress values as calc by Nastran for skin/panel (2D) elements directly.

Rgds,
- VN

 

[rb1957]

it sounds like you're trying to find the shear flow along a line of rivets ?

This is pretty straight forward in the case of a stringer on a skin panel ... calc the change in endload along the length of the stringer, no? Personally, I think I'd take the .f06 results into excel and manipulate there.

another day in paradise, or is paradise one day closer ?

 

[VN1981]

RB,
The main intent of my question is to understand if this method is still applicable today? Like I'd mentioned in my OP, we seem to have missed this step.

Outside of Ian Tiag & one more document, I've never come across any references to carrying out the addition of shear load contribution from 2D to 1D elements.

I have no experience on this excpet understanding the procedure. So, I wanted some inputs on how well this procedure is used in industry.

Thx

 

[rb1957]

can you upload Ian's doc to engineering.com ?

another day in paradise, or is paradise one day closer ?

 

[VN1981]

Sure RB,
Here is the copy of Ian Tiag. Its about 4 Meg in size.

http://files.engineering.com/getfil...5a1916&file=Ian_Taig_-_Aircraft_FEM_notes.pdf

Note: I tried to upload my copy several times to Engineering.com, but failed. Luckily, found the above link in another thread where the poster 39minuteman had uploaded a copy of Ian Tiag. Credit goes to him.

Regards

 

[VN1981]

Good morning folks!

RB & others, I hope you guys had a chance to review Ian Tiag's notes on 2D elements' shear & end load contributions. Hopefully, I will get to find out if that procedure is still in use today

RB19857, I had a question about delumping of loads based on effective width. You've mentioned that current practice is to model the skin as fully effective and then delump the stiffener loads of FEM loads. I was talking to a colleague of mine & an interesting question came up.

Lets say that we are modelling a wing with stringers. As you've mentioned, the skin is modeled as fully effective and stiffeners are modeled as 1D elements. The whole FE is modeled as integrated in global representation i.e. no fasteners. Now if the object of interest is how much compressive loads is seen in top skin at any 2 sections along the wing span, the skin elements also carry internal axial loads since they are effective. Now if you are delumping the stiffener loads considering the effective area of skin, does this delumped load add to the existing skin load output from FEM? I dunno if I have made my question clear. Lets say that argument purposes, our wing section between rib stations consists of 1 skin panel & 2 stringers (ignore the spar caps for now). If the total internal axial load is Pa and the load carried by skin & str is Pskin & Pstr respectively. From FEM output, once can calculate Pskin & Pstr respectively. Now let us designate the delumped skin load Pdskin. So does the new skin load become Pskin + Pdskin.

Hope the above question is not stupid

Thanks

 

[rb1957]

if you were dealing with a fuselage, I'd say take the GPFB force at a node and proportion it between the stringer area and the effective skin area (ie 30*t^2) ie Pstr = Ptotal*Astr/(Astr+30*t^2).

A wing however is notoriously sensitive to the offset of the stringer from the skin plane; so I'd probably consider an entire rib section at once.

This assumes that you modelled the skins as fully effective. The olde school approach would have modelled the effective skin as part of the stringer and modelled the skins as shear panels.

another day in paradise, or is paradise one day closer ?

 

[VN1981]

Hi RB1957,
Thanks for detailing the procedure to delump loads but I was aware of the formulation.

I am afraid my question was posed very clear. I have made a small presentation (< 5 slides) in hope that my question is clear. I would appreciate if you take a look when you get a chance & respond.

Thanks

 

[rb1957]

ok, slide 5 ... I don't think we know Pstr from GPFB. Sure GPFB will say the load in the stringer element, but this is the load with fully effective skin. From GPFB all I think we can use is the total load, and proportion that between the stringer and skin, A total = Astr+30*t^2.

But like I said, I think this is not good practice for a wing, as the stringer is offset from the skin. For a wing, I'd use the total moment and endload (summing the GPFB from each node). For a wing I would off-set the stringer from the skin, either using an element offset, or a dedicated stringer node.

another day in paradise, or is paradise one day closer ?

 

[VN1981]

RB,
Thanks for the response.

>>But like I said, I think this is not good practice for a wing, as the stringer is offset from the skin
The str element (BEAM) used offset.

If I have understood your response correctly, GPFB provided by STR element may not represent the actual load in STR since skin is modelled as fully effective. In order to calc the Str load, it needs to be delumped.

After delumping, the contribution from skin will go on to add to existing GPFB load from 2D elements, correct?

Sorry if I am asking the same questions over & over again. Just want to get as much clarity as possible before I take it forward. I appreciate your patience on this thread.

 

[rb1957]

for a wing I wouldn't do what I think you're proposing. I think you should extract a complete wing section load (endload and moment) from the GPFB (by summing loads from all nodes), and apply this to your real geometry ... stringers modelled at their centroid, 30t effective skin

another day in paradise, or is paradise one day closer ?

 

[VN1981]

Hi RB1957,
I have been thinking about your explanation and I am sorry being so thick, but I am still having some issues. Perhaps what I really need is a much more elaborate explanation which may not be possible on a forum like this due to obvious constraints.

Apart from Ian Tiag's FEM notes, is there any other material in open source which provides much more comprehensive explanation? I would appreciate if you could point it out to me.

Thx in advance,
- VN

 

[rb1957]

sorry can't really "dumb down" my previous post.

1) from GPFB extract the total load in the FEM at the node,
2) for each node on the cross-section,
3) sum these loads, total endload, shear, moments, torque on the cross-section.
4) carefully create the section properties of your cross-section ... skin and stringers.
5) apply the overall loads (from 3) to the detail section (4) and calc stresses.

another day in paradise, or is paradise one day closer ?

 

[VN1981]

RB,
Thanks for the explanation again. I think you have hit the nail on the head this time. I think I may have got it.

So the real geometry as mentioned in (4) would be Stringers plus 30t effective skin right?

 

[santoshaero]

Hi RB,

This is with respect to your post on 16 Nov 15 19:34.

Is there any difference to modelling of skin stringer in fuselage and wing skin, as you repeatedly mention the skin offset. Is it because of thicker wing skin compared to fuselage skin?

 

[rb1957]

the depth of the fuselage is much bigger than the depth of a wing, so stringer offset is less important for a fuselage (if you want, the change in bending stress over the depth of a stringer is less (much less) in a fuselage than in a wing, so the effect can reasonably be ignored in a fuselage (model stringers as 1D elements on the skin surface).

another day in paradise, or is paradise one day closer ?