Vanishing Jacobian

[Daniah Aljubaili]

Hi Neks,

I have the following error below when I created a 3D channel with a cylinder using fluent. I have specified the boundary conditions correctly but I still get a vanishing Jacobian error as below. Why does this happen? Is it because of clashing elements? or what is the main cause of this? I have tried a simple channel and that seemed to work fine under the same usr file but I cannot understand why this error occurs when I use my model.

  6  ERROR:  Vanishing Jacobian near      6th node of element     19921.
   1.5571460688003769E-002  -2.3621147210481168E-004
    6 xyz:   2.36258E+01  -4.89235E-01   3.87716E+00
    6 xyz:   2.34586E+01  -5.29548E-01   3.81090E+00

    6  ERROR:  Vanishing Jacobian near      9th node of element     19973.
   1.0172731881245961E-002  -1.3046785157854330E-002
    6 xyz:   1.72868E+01  -4.95416E-01   4.32160E+00
    6 xyz:   1.77396E+01  -4.94975E-01   4.12284E+00

    6  ERROR:  Vanishing Jacobian near      5th node of element     19974.
   2.7368113558374880E-003  -1.0966939019799315E-002

    7  ERROR:  Vanishing Jacobian near      6th node of element     16154.

Please anyone help!

[Tanmoy Chatterjee]

The elements 19921, 19973 have negative Jacobians. The elements are not convex hexahedral, it seems, and you need to remedy. Did you check/visualize those elements in your meshing software and verify that they are severely distorted? 

Best,
Tanmoy

--
You received this message because you are subscribed to the Google Groups "Nek5000" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+unsubscribe@googlegroups.com.
To post to this group, send email to nek...@googlegroups.com.
Visit this group at https://groups.google.com/group/nek5000.
To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/CAN_cHZtP6jzGOCvgpn00DYKomSgGWVBJWq5eBvPNhVJ4w4rgjQ%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

--

Best Regards,

Tanmoy

[Tanmoy Chatterjee]

Hi Daniah,

From what you send it seems some of the mesh are highly skewed. Probably your 2D O grid had acceptable skewness, so it worked. Honestly, I can't think of a way to fix this Nek, unless you remesh it. You could have used the mesh smoother, in Nek5000 v19, if you did not negative jacobians. You could try to see if there are any mesh smoothing tools in fluent to improve the soreness of the meshes. 

Best,

Tanmoy

On Wednesday, July 10, 2019, Daniah Aljubaili <daniah.a...@gmail.com> wrote:

Hi Tanmoy,,

Yes I have visualised and it seems all the elements are distorted (find below). My mesh created on fluent was the same as a case I made for 2D (O-grid)  and works fine its when it's 3D it's when I get this error. Is there another way to remedy apart from fixing the mesh through fluent because I do not think there is anything wrong ?

To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/CAPiUqMimcua_GuPFcRzc7qSt5%2BW43%3Dqg0X85s-hAA-8%2BxLJXPw%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

--

Best Regards,

Tanmoy

[nkul...@ncsu.edu]

Hello,

I am trying to run a case with a structured hex20 mesh that has previously worked with Nek19 without errors. Currently I'm trying to run it on a new system and have loaded a different set of compilers but it gives me vanishing jacobian errors for a lot of elements during runtime. There is no error while loading the mesh through exo2nek and genmap and no error with makenek either. What could be the possible cause for the vanishing jacobians?

Thanks in advance.

Nilay

To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+u...@googlegroups.com.

To post to this group, send email to nek...@googlegroups.com.
Visit this group at https://groups.google.com/group/nek5000.
To view this discussion on the web visit https://groups.google.com/d/msgid/nek5000/CAN_cHZtP6jzGOCvgpn00DYKomSgGWVBJWq5eBvPNhVJ4w4rgjQ%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

--

Best Regards,

Tanmoy

--
You received this message because you are subscribed to the Google Groups "Nek5000" group.

To unsubscribe from this group and stop receiving emails from it, send an email to nek5000+u...@googlegroups.com.

To post to this group, send email to nek...@googlegroups.com.
Visit this group at https://groups.google.com/group/nek5000.