"Error: Vector length = NaN"

[bjorn_fallqvist]

I'm running an MES with large deformations (plastics) and contact. It is basically a bag which I apply a pressure on (shell elements) and allow to expand. However, there is contact invovled, and it seems that depending on some contact settings and the material properties, sometimes I get the error shown below:

Time        Percent     DT       L Iter. Residual
   0.000000   0.00     0.100000  1   0   0.0000E+00
   0.100000  10.00     0.100000  1   1   1.0000E+00
   0.100000  10.00     0.100000  1   2   9.7122E-02
   0.100000  10.00     0.100000  1   3   9.5230E-01
   0.100000  10.00     0.100000  1   4   9.9943E-01
   0.100000  10.00     0.100000  1   5   1.0000E+00
   0.100000  10.00     0.100000  1   6   1.0000E+00
   0.100000  10.00     0.100000  1   7   1.0000E-02c
 ERROR: Vector length = NaN           Vector= NaN           NaN           NaN          
 Possibly bad normal vector or guiding vector direction.
 ERROR: Vector length = NaN           Vector= NaN           NaN           NaN          
 Possibly bad normal vector or guiding vector direction.
forrtl: severe (37): inconsistent record length, unit 33, file C:\Windows\system32\fort.33

Image              PC                Routine            Line        Source             
arun4.exe          0000000140E176C4  Unknown               Unknown  Unknown
arun4.exe          0000000140E12EDD  Unknown               Unknown  Unknown
arun4.exe          0000000140DB580E  Unknown               Unknown  Unknown
arun4.exe          0000000140D9C73D  Unknown               Unknown  Unknown
arun4.exe          0000000140D9BBDD  Unknown               Unknown  Unknown
arun4.exe          0000000140D7661E  Unknown               Unknown  Unknown
arun4.exe          0000000140005282  DOHEAD                     72  dofile.for
arun4.exe          0000000140441170  EXITN                      94  yexitn.for
arun4.exe          0000000140760AE6  CLEANQUIT                 148  cleanq.for
arun4.exe          0000000140717927  SETRST                     72  shnorm.for
arun4.exe          00000001406BB7EF  SHELKM                    893  shell1.for
arun4.exe          00000001406BA403  SHELKM_WRAP                24  shell1.for
arun4.exe          00000001403A6165  SHELL                     273  shell.for
arun4.exe          0000000140092FAE  ELEMNT                     82  zelemnt.for
arun4.exe          000000014023004A  UNBLDF_BODY               974  xzutil1.for
arun4.exe          000000014022DB52  UNBLDF                    665  xzutil1.for
arun4.exe          0000000140691E18  EQUIT4                    608  xequit4.for
arun4.exe          00000001407FCBC3  NLLOOP                     83  equilibr.for
arun4.exe          00000001407E540C  VSS_IMPL                  615  vssimpl.for
arun4.exe          000000014039D7FA  FMAIN                     600  yapak4.for
arun4.exe          0000000140493A7D  MAIN__                    327  fallocn.for
arun4.exe          0000000140E1FC84  Unknown               Unknown  Unknown
arun4.exe          0000000140DF2A75  Unknown               Unknown  Unknown
kernel32.dll       0000000076E3F56D  Unknown               Unknown  Unknown
ntdll.dll          0000000076F73021  Unknown               Unknown  Unknown

I assume it has something to do with calculating the normal vector for the contact force, but I'm at a loss of how to handle the problem. Any suggestions?

Thanks

Björn

[bjorn_fallqvist]

I also seem to be getting this problem when I refine the mesh.

[S.LI]

It seems this error is from normal direction calculation for shell element.

One of the possible reasons could be badly distorted shell element due to plastic material model.

I really appreciate it If you can share your model.

BTW: what's the version of your MES?

[bjorn_fallqvist]

Yes, that's what I was thinking. I'm uploading the archive file now.

Do you think it could be due to my definition of the normal point? I always assumed that it was enough to create a point on the right side of the element, and then this would be understood by other elements of the part, even if the surface happens to be curved. Is this a misunderstanding on my part?

[S.LI]

After reducing the step size, the model runs well.

Thanks for sharing your model, and I think the code should at least handle the exception in a better way.

[bjorn_fallqvist]

Ok, thanks. Regarding my other question, about the orientation of shell elements, did I misunderstand the way it was supposed to be defined?

[S.LI]

Yes, what you did is good enough. The shell element normal here is for setting the correct direction for contact and/or pressure load.

The Nan error is not related to this setting.

In a general nonlinear analysis, codes try to find the solution with iteration methods. Sometimes, these estimated solution is so bad that elements become distorted. In this situation, codes either stop or automatically try smaller step size.

In your model, codes try to quit but trigger some I/O error.

So your model is totally fine and small step size will help codes converge to the accurate solution quickly and easily.

[bjorn_fallqvist]

Hi,

Ok, I am getting this error message constantly in one of my models including shell elements (hyperelastic ones). Is the only way to get around this to reduce step size? That would make the runtime ridicolously long. I will try to apply the load slowly at first (to allow unfolding, where most of the shell normal directions probably change) and then apply it more drastically.

Can we hope to see this fixed in future versions of Algor?

[zhuangs]

One defect has been reported.  And the fixe will be added in the next new release (Autodesk Algor 2012):  MES will reduce time step with message rather than crash in this case.

-Shoubing

[zhuangs]

The fix in Autodesk Algor 2012 has been done.  The code will automatically reduce time step with outputing the follow message:

ERROR: Vector length = NaN Vector= NaN NaN NaN

Possibly bad normal vector or guiding vector direction.

** Time step reduction due to: bad normal vector or guiding vector direction

However, the user needs to make sure the guiding vector direction is correctly set.

[bjorn_fallqvist]

That is good to know, thanks.