THERMAL WITH RADIATION - SOLUTION OUTPUT - VIEWFATCOR SOOMTHING

fabrypirofabrypiro Member Posts: 4

Hi guys,

I am carrying out a transient thermal analysis with exchange by radiation. From an Ansys tutorial I learned that to validate the accuracy of the results you need to check the "Viewfactor Smoothing" and make sure that the first 4 values are the same and the errors are very small (at least e-5).

I need some more information on the meaning of these values.

I also noticed an unexpected phenomenon: by improving the mesh the difference between the 4 initial values increases:


MESH SIZE 50 mm

MESH SIZE 10 mm


Could you help me ?


Thanks in advance,


Fabry

Comments

  • SheldonISheldonI Posts: 112Ansys Employee

    Hi Fabry,

    If you have a perfect (i.e., completely closed) enclosure, the view factors for each element face should add up to 1.0. Depending on your geometry, an element may not properly 'see' another element. Thus, by either increasing the Hemicube resolution (under "Analysis Settings") or by increasing the mesh density (smaller elements), you can achieve better viewfactor calculations.

    Your values seem close, so even your mesh size of 50 mm should be fine. View factor 'smoothing' simply scales the viewfactors so that they do add to 1.0. For example, because of the coarse mesh and geometry, the viewfactor could be 0.99 instead of 1.0, but by 'scaling' the values, we can get 1.0. It should never be used if the viewfactor sum is way off, but your case seems fine.

    If you are interested in details of the view factor calculation technique, please refer to the Mechanical APDL documentation, Theory Reference, Section 6.4.2. "View Factor Calculation (3-D): Hemicube Method".

    In summary, the error doesn't need to be 1e-5 to be numerically accurate - your 50 mm case should be fine. And if you still want better accuracy for viewfactor calculations, instead of decreasing the element size, you can increase the hemicube resolution. However, either approach will add some computational cost/overhead.

    Regards,

    Sheldon

Sign In or Register to comment.