I am working with a volume tetrahedral mesh, where Youngs modulus is temperature dependent. Each node temperature is defined separately and I know the highest temperature in the model. When outputting nodal temperature (NT11), I get exact correspondence. For linear C3D4 elements, when outputting temperature at the integration points (TEMP), the lowest value is slightly higher (and highest slightly lower) than NT11, as I had expected.
However, for quadratic CD10 elements, I noticed that minimum and maximum element temperatures surpasses the min and max nodal temperatures.
Since the element stiffness is temperature dependent, I am curious about whether element stiffness potentially becomes higher (or lower) than expected.
Are anyone able to explain the reason for this?
The integration point temperatures are calculated from the known nodal temperatures by the shape functions.
Especially for distorted mesh (often 10 node tetra) you will get integration point temperatures that are outside the 'physical boundaries'. Say that you have convection on both sides of a meshed wall, 20 degree on the inside and freezing -20 on the outside. Then your integration point temperatures can be above 20 or below -20. And it depends on the element type and quality.
For that reason it is common practice to specify the temperature dependent material data with margin, for instance by adding -30 and +30 to the data, and putting the same values as the -20 and 20.
Nils Wagner, I have appended the .inp files for the two coarsest meshes, but the problem is the same for finer meshes as well. The "dens...txt"-files are lists of nodal temperatures which the .inp extracts data from. "material.txt" is the temperature-modulus table.
Kenth Nilsson, I do have some distorted elements. The meshing is performed in Materialise 3-matic, and I haven't figured out any good way to treat this problem.
Your explanation makes good sense, although my "temperatures" are actually treated as bone density at the nodes which is why values might not change evenly in space.
Nevertheless, since Abaqus calculates modulus of temperature values outside the defined spectrum as the same as spectrum endpoint, this shouldn't really create any problem. I am only concerned that elements with nodal temperatures within the spectrum might occur stiffer/softer than they should be.
- Badges
- Science method