CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   Floating point error (https://www.cfd-online.com/Forums/main/13621-floating-point-error.html)

mvee June 12, 2007 02:49

Floating point error
 
hi,

I found the message during solution as

Error: Floating point error: invalid number Error object: ()

can anybody tell me about this?

I have tried by

1) Reducing mesh size 2) Changing turbulence model 3) changing turbulence intensity in boundary condition panel 4) Not solving all the equations simultaneously(solve-control-solution panel). Solving the equation one by one or in group of two equations.

I spend lot of time to solve this problem. Can anyone have faced the same one and at the end got the solution, please reply me.

Your failure in this case also may give me the way towards the solution.

thanking you in advance and please reply whatever you know in this respect.

regards, mvee

rlindsey June 12, 2007 11:26

Re: Floating point error
 
Hi, I had some of those when I was experiementing with my hardware. I am overclocking my cluster and when I pushed it up too high it will throw those, even if the setup appears stable on some benchmarking tests. I just backed down the speed a little and it's fine.

Are you OC'ing your machines? How old are they? If they get dust in the heatsinks sometimes they will get hot and let an electron go the wrong way... Have you run coretemp or anything to see if your processors are heating up?

I'm sure there are other causes for this but that's the one I ran into.

Markus Lummer June 13, 2007 00:56

Re: Floating point error
 
http://www.cfd-online.com/Forum/flue....cgi/read/4193

mvee June 13, 2007 04:04

Re: Floating point error
 
Hello Lummer,

I have tried the solution that you have given through the link. I have checked the skewness it is in the range of 0.14. And also i have changed the grid size, reduces the size of grid. About initialization i am confident because i ran the problem with different dimension and it is converged. I am using PISO scheme for Pressure velocity coupling and PRESTO scheme for pressure discretization. So i kept all URF as 1. AT the same time i checked it by changing the URF. Still floating point error is there.

[in your link the author use the word as "numerically stiff situation". What does this mean?]

can you suggest any different scheme?

thanks for your response

regards, mvee


All times are GMT -4. The time now is 04:30.