Stress linearization

[jbw9]

All,

I am begining to learn the process of stress linearization (in general). Do y'all have any suggestions in terms of references I should start with? I will need to use this technique to perform some analyses on 3-D FEM(s). After reading some of the information on this forum, it seems the largest hurdle is figuring out how to seperate the membrane and bending stresses from the outputs available in the finite element code being used. I will be using ABAQUS which apparently has this capability built in. However, Im hoping to extend this methodology to the software I am most familiar with which is NeiNastran. Has anyone had experience with stress linearization in NeiNastran? If so, I would love to discuss.

Best to all.

 

[prex]

Concerning ABAQUS's capability of separating membrane and bending stresses from the balance stresses, well that's not really difficult, but be careful, as some softwares do not do this operation correctly: this was referred to in this forum also by others (I know only about ANSYS).
Consider that this is just a mathematical operation, you'll be able to do it by hand or to automate it with some macro language built in your software.
However what you should really study in depth, because it's a much more difficult task that cannot be automated into a software, is how to separate primary from secondary stresses, and peak stresses from the total stresses. The only way to somewhat automate this is by plastic analysis, but if you want to stay elastic, then it's a matter of experience and reasoning (and 3D analyses make this task much harder).
There are many threads in this forum about the subject: you could start from there, then try some worked examples.

prex

http://www.xcalcs.com

: Online engineering calculations

http://www.megamag.it

: Magnetic brakes and launchers for fun rides

http://www.levitans.com

: Air bearing pads

 

[jbw9]

Thanks very much for the response prex. I can see where determining which contribution to the the stress components is due to secondary/primary stresses would be somewhat subjective, or based on engineering judgement I should say. Eventually, I will have to be able to make that judgement. Am I correct in saying that when performing stress linearization (in FEA) you have to perform the following steps initially (say we know all the stresses are primary):

1) locate a stress classification line (SCL) - of course, you need nodes located along this line.

2) for extract the "nodal" stresses for all six stress components (3 normal and 3 shear?)

3) manipulate these stresses to seperate the membrane and bending stresses (i.e. membrane is the constant part accross the cross section and bending is the linearly varying part across the cross section).

Please correct me if im wrong. I also noticed that the principal stresses are used in some way which i did not understand.

 

[prex]

1)You don't necessarily need the nodes along the SCL: the software may be capable of averaging the stresses to obtain them at any point. But of course for manual manipulations it is very useful to have the nodes on the SCL
3)There's not so much to manipulate: by averaging each one of the six components along the SCL (note: the averaging procedure must account for the positions or distances along the SCL of the used nodes), you obtain the six components of the membrane stress and similarly for bending (averaging the moments of each component about the middle of the SCL).
You'll generally combine those 6 components into a single quantity to check the stresses against a failure criterion. The Tresca criterion requires determining the 3 principal stresses to obtain the largest difference of them taken 2 by 2.
To be noted also that the bending stress alone is not really useful, as the check will be on the membrane+bending stress. Now by the procedure above you obtain two different membrane+bending stresses, one at each end of the SCL, and of course both need be checked.

prex

http://www.xcalcs.com

: Online engineering calculations

http://www.megamag.it

: Magnetic brakes and launchers for fun rides

http://www.levitans.com

: Air bearing pads

 

[TGS4]

jbw9 - I would highly recommend that you review the new (2007 and newer) edition of Division 2, Part 5. The instruction there about how to perform the linearization are top-notch. If something there is not clear, come back here with additional questions.

 

[jte]

I guess my thought is... Why take an inherently nonlinear program such as Abaqus, model something, then run it linearly and got through allthis effort? Why not just use the nonlinear side of Abaqus and be done with it? Like TGS4 pointed out, the new Div. 2 is worth reading. Though I think the linear parts of Part 5 could just be skipped...

jt

 

[TGS4]

jte - agreed. I do just about every analysis now as elastic-plastic. Just not worth the effort anymore to perform the stress linearizations. Too much work to linearize, categorize, etc.

 

[corus]

The basic idea of stress linearisation is to provide you with the membrane, primary/secondary bending stress and the peak component stress.

The peak stress is used to calculate the fatigue life of a component for which SN curves are normally used. These curves are based upon calculated (linear) stresses within a specimen, and hence must be compared with linearly calculated stress results from the FE model. Using Abaqus with non-linear (post yield) stresses would be wrong for this case.

corus

 

[TGS4]

corus - please explain how using elastic-plastic analysis for fatigue would be "wrong"? It's permitted in ASME Section VIII, Division 2 (2008), Part 5, Article 5.5.4. Is there something that you find incorrect in the formulation of Article 5.5.4?