Cancel
Showing results for 
Search instead for 
Did you mean: 

Percentage Error - Uncertainty Analysis

Builder
Builder

Hello everyone,

 

My question might be out of forum.

 

However I am just seeking advice

 

As I have performed experiment as well as FEA and CFD (Using FEMAP and TMG) and now I want to see the percentage error.

 

Does anybody know how to do uncertainty analysis, any software recommendation or any advice would be very helpful for me?

 

Many thanks in advance for your support and help.

3 REPLIES

Re: Percentage Error - Uncertainty Analysis

Solution Partner Phenom Solution Partner Phenom
Solution Partner Phenom

Dear Varum,

Most of the answer is in your hands: you have an element library & different analysis types, and also you can play with mesh densities, then use all of those variables an compare results.

  • Not only mesh the problem with 3-D solids, also use 2-D mesh or if available 1-D beam elements, solve and compare results.
  • At the same time, play with mesh size and compare result: beam elements are discrete elements, the answer is the same not matter the mesh density used, but for shell & solids the mesh density is critical. The target: you need to arrive to a mesh undependent model, ie, the answer should not depend of the mesh density. The solution is convergent when the difference between two analysis whit double DOFs is not bigger than a 5% or so.
  • Also, play with analysis types, and compare results: linear static analysis give an answer, yes, but real life is always nonlinear (by geometry, material & BCs), then linear static is a simplification of the real life. If the nonlinearities are not severe, then a linear static is enough, but you need to judge the level of nonlinearity of a problem, the way is run both analysis & compare results, OK?.

Best regards,
Blas.

Blas Molero Hidalgo, Ingeniero Industrial, Director
IBERISA • 48004 BILBAO (SPAIN)
WEB: http://www.iberisa.com
Blog Femap-NX Nastran: http://iberisa.wordpress.com/

Re: Percentage Error - Uncertainty Analysis

Builder
Builder

Thanks you Blas for the answer.

I understood what you described above.

But my points is how to perform it.

In relation to your above statements, well, my model is in STL Format so cannot vary the mesh size and elements. So cannot do much variation in mesh. But I am using solid linear tetra 4 noded elements

However, I have performed crushing analysis on porous material by choosing elastic – plastic material model.

So now, I have experimental data which consist stress vs strain value and FEA von misses stresses, strain, displacement results.

Now, I wanted to compare experiment vs simulation results to see how much percentage of error occur.

My points is how to perform it, Shall I use basic uncertainty equations or statistics or is there any other way to do.

Re: Percentage Error - Uncertainty Analysis

Solution Partner Phenom Solution Partner Phenom
Solution Partner Phenom

Hello!,

First at all: TET4 is a "awful" element type, is a constant strain & constant stress element, each point internal in the element has the same stress & strain, then its use should be avoided always, please consider using high-order TET10 elements, in FEMAP you can convert your mesh using MODIFY > UPDATE ELEMENT > LINEAR/PARABOLIC ORDER command. In severe nonlinear analysis the use of high-oder elements is mandatory, and preferable HEX elements instead TET ones.

Best regards,
Blass. 

Blas Molero Hidalgo, Ingeniero Industrial, Director
IBERISA • 48004 BILBAO (SPAIN)
WEB: http://www.iberisa.com
Blog Femap-NX Nastran: http://iberisa.wordpress.com/