Why specify a maximum strength non-linear or linear analysis? Which is correct for rectangular tubes 3

[ecspt]

Hi All,

I am an entry level engineer, and I am doing FEM analysis in the company I work in. I am analyzing rectangular tubes in static analysis. I am doing a linear analysis, with linear material properties. However, if I do a non-linear analysis, the results are quite different. The rectangular tube has length of 1 m, section height of 0.3 m, section width of 0.1 m and thickness of 0.002 m. I am applying a concentrated transverse load of 1500 N on the center of the top area. How do I know if I shall do linear analysis or non-linear analysis???? which one is correct?


thanks,
Mark,

 

[TGS4]

Of course you will get different answers. What criteria are you evaluating the results against? Is it one based on elastic analysis or one based on non-linear analysis? Note that normally, a non-linear analysis is performed as an LRFD (I'll leave that to you to Google).

 

[inline6]

post some pics or something of the two answers, "quite different" doesn't mean much.

if it wasn't so expensive to do non linear we would do it for every analysis as everything is non linear in the real world, its justwith some problems the non linearity is small enough that linear gives a good enough approximation

 

[ecspt]

Hi,

Thanks for the reply,

I am attaching pictures of the results in both cases. My question is: Shall I do a non-linear analysis or linear analysis? Which results are correct? By which analysis the results are closer to an experimental test, if it is done in the same conditions?

The files are:

Linear

http://files.engineering.com/getfil...0-4a3f-8868-57a5a0c45bb7&file=linear_viga.png

Nonlinear

http://files.engineering.com/getfil...d25-8f25-ca82bce353f4&file=nonlinear_viga.png

Best regards,
Mark

 

[Brad805]

Since you have a 2mm wall thickness, my first concern would be if the point load is representative of the actual load application? The displacement is highly localized, so unless you have a perfect point load I would be suspicious of the deformed shape.

 

[TGS4]

Your plots are pretty useless in the discussion. Are the stresses in excess of yield? Are you expecting buckling (or similar phenomenon)? Are you trying to perform a design or match a test?

 

[ecspt]

Hi TGS4,

Thanks for your reply

No, actually stresses both by Von Mises and Normal stresses criteria are not passing yield. My concern is if the geometry by itself suffers non-linearity due to buckling. If I do a non-linear analysis in ANSYS it does not say "MAX PLASTIC STRAIN STEP", even in the latest substeps. I did this with both SHELL63 (Elastic element) and SHELL43 (Plastic element). Ansys says in the help that NLGEOM,1 shall only be used when the strain is higher than a few percent. I have maximum strain, both in intensity and by the Von Mises criterion lower than 1%. Am I right if I assume that the material does not reach the plastic domain, and therefore a linear analysis can be a good approximation of the results? This should mean that I am doing a non-linear analysis and considering non-linear effects when they are not present. This explains the results divergence shown in the charts. Do you agree with what I am saying?


thanks,
regards,
Mark

 

[TGS4]

As you are likely aware, there are two types of non-linearity: material and geometric. If your stresses do not exceed yield, then you are likely not feeling with the former. However, the latter can occur regardless of the magnitude of the stress; classic case in point is elastic buckling.

Another classic example is internal pressure on a flat plate. Initially, the stress in the plate is pure bending. However, once a small amount of deformation occurs, some membrane stress develops. At its extreme, the flat plate can be "hydro-formed" into a hemispherical head.

You need to decide for yourself if the manner in which you structure is loaded could have geometric nonlinear effects, or not.

 

[inline6]

it looks like the non linear result is showing the stress stiffeneing and thats why what looks like deflections are lower for the non linear as it captures the true stiffness.

i think the rule of thumb is if the face deflects more than half the wall thickness then use non linear to capture stress stiffeneing