Highlighted

Element Quality Sensitivity

Valued Contributor
Valued Contributor

Can anyone explain the relative sensitivity of Nastran to each of the element quality metrics (i.e. Jacobian, skew, aspect ratio, etc.)? I am aware of how they each affect the quality of my results, but I am curious if anyone knows how significant those effects are, relative to each other. For example, will small changes to the Jacobian have more significant downstream effects than small changes to element aspect ratio?

3 REPLIES 3

Re: Element Quality Sensitivity

Siemens Valued Contributor Siemens Valued Contributor
Siemens Valued Contributor

 

Hi, I work for Siemens PL and I can assure you that in respect to element quality loss due to distortion, NX Nastran is no different than any other finite element code. The sensitivity to distortion is not solver specific but dependent of the element formulation. E.g. a highly distorted element is more suitable for a reduced integrated element formulation than for a fully integrated enhanced strain formulation element.

 

But to you point: There is no such thing like a "look up table" on quality loss depending on certain distortions. All elements in NX Nastran comply with the so called "patch test": whatever the distortion is, the result will be exact under constant, unidirectional strain. This patch test is one prerequisite for any "valid" element formulation. This shows that the sensitivity to distortion is very much also depending on the loading. You can without problem have an aspect ratio of 1:100 for a hexahedron element as long it is otherwise undistorted and the strain gradient is "small" in the direction of the long side. Please note that a parallelogram/parallelepiped is an "undistorted" element with no element quality loss. "Dangerous" are trapezoidal elements, there I would go for underintegrated elements to avoid locking.

 

Sorry, I would have liked to give a more precise answer, but there is no..... The ultimate check for mesh quality is a refinement: If the results change significantly after mesh refinement, the initial mesh was too coarse or too distorted. With mesh refinement, a distorted trapezoidal element will eventually move towards a mesh of parallelograms...

  

Good luck and happy festive season!

 

Martin

Global Simcenter Portfolio Development

Re: Element Quality Sensitivity

Valued Contributor
Valued Contributor

Thanks, Martin! I had a fear it would be an "it depends" answer, but this is very helpful.

Re: Element Quality Sensitivity

Phenom
Phenom

To add to the previous post

Tests were done years ago and reported with different elements quality on very simple model eg a beam in bending with a few elements. Check Nafems or this link EFFECTS OF ELEMENT DISTORTIONS ON THE PERFORMANCE OF ISOPARAMETRIC ELEMENTS

Production: NX.CAE 10.0.2.6
Development: VB.NET (amateur level !)