Floating point error! - Overflow

A discussion forum for old Hydrus-2D & Meshgen. Happy Posting!
Download PDF document with all topics (February 8, 2008, 1.9MB):
http://www.pc-progress.com//Downloads/F ... on_H2D.pdf
jsamson
Posts: 16
Joined: Thu Feb 02, 2012 7:14 pm
Location: USA

Floating point error! - Overflow

Post by jsamson » Tue Feb 07, 2012 7:53 pm

I keep getting this message, but it only occurs with certain initial conditions. Can you please let me know why this may be?

Jirka
Posts: 4712
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Post by Jirka » Wed Feb 08, 2012 12:51 am

YOu probably need to adjust spatial or temporal discretization when you use more extreme (e.g., drier) initial conditions. Check out these notes:

http://www.pc-progress.com/Documents/No ... zation.pdf

J.

jsamson
Posts: 16
Joined: Thu Feb 02, 2012 7:14 pm
Location: USA

Post by jsamson » Wed Feb 08, 2012 4:22 pm

Alright, so I tried to increase the spatial discretization, given that my temporal discretization seems like it matched up with the info contained in the link you shared, but now other issues are coming up. I have layers that are 2.5cm thick, and I am trying to cut these in half but the software is giving me prompts like: "nodes 68, 118 have identical coordinates...". However this is incorrect. I selected each node separately and their coordinates are not identical. The fact that I can see them in space as distinct points also reinforces that their coordinates are not identical. Now the problem becomes how do I increase spatial discretization when Hydrus provides me contrary errors as well as does not allow me to translated existing lines?

Mirek
Posts: 406
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Wed Feb 08, 2012 5:01 pm

I do not know how large is your domain and therefore it is difficult to guess what the problem is. For example, if the size of your domain is 10km and you have two geometrical points (not FE-mesh nodes) spaced 2 cm, you can get this error message. If you send me the project, I can look at it (ZIP it and send it to support@pc-progress.cz).

jsamson
Posts: 16
Joined: Thu Feb 02, 2012 7:14 pm
Location: USA

Post by jsamson » Wed Feb 08, 2012 9:11 pm

I sent you the model. Please let me know what you think when you get a chance.

Mirek
Posts: 406
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Thu Feb 09, 2012 9:55 am

Although you posted your question to this forum (old Hydrus-2D), your project was actually created in HYDRUS 2D/3D version 1.x.

Dimensions of your domain are 250 x 7 [m] and there are many horizontal layers with thickness of 1 cm (see the attached figure). I do not know what you want to simulate but I would say the horizontal flow will be very slow (or negligible) and there is a question if this model is reasonable. May be Jirka Simunek will add a comment here.

Although 1 cm layers in such a long domain are close to the model resolution limit (which is 2mm), it should be possible to create such domain and generate a FE-mesh. I tested it in version 1.x and there indeed were some problems with geometrical calculations and the program was displaying incorrect warnings about intersecting lines. I remember that we improved some geometrical calculations in version 2.x and therefore I opened this project in HYDRUS 2.01. I was then able to work with your domain without any problems (see the generated FE-mesh). However, I do not think that this FE-mesh is very good for HYDRUS calculations. I would not use additional horizontal lines (as you have done) but I would rather use FE-mesh stretching in X-direction in combination with some local mesh refinements.

Mirek

Image

jsamson
Posts: 16
Joined: Thu Feb 02, 2012 7:14 pm
Location: USA

Post by jsamson » Thu Feb 09, 2012 3:56 pm

I thought we are using the old version of hydrus, because last week when I asked you the price of an upgrade you told me we have the old version and it would cost the full amount. I sent you info on our license through email. I get your system confused and I did not know there was a difference between the two. As for my model, my soil layers range from 5cm to 30cm+. I was trying to cut these 5cm layers into smaller layers to increase the spatial discretization as was detailed in the document Jirka shared with me. In my system, under flooding conditions research has shown that horizontal flow will dominate; especially given the high conductivity of some of the thicker layers. Therefore I don't think the question is of whether or not my model is reasonable, I think the question should be whether or not I am going about increasing the spatial resolution correctly in order to not deal with the high gradient conditions. About the incorrect warnings, what I am to do about that?

Mirek
Posts: 406
Joined: Sat Mar 16, 2002 4:00 pm
Location: Czech Republic
Contact:

Post by Mirek » Thu Feb 09, 2012 5:23 pm

This starts to be interesting:

1/ Last week, when we were discussing the upgrade and I asked you about your current license, you sent me Dialog A (see the attached figure). This is a dialog from old Hydrus-2D & Meshgen, which was distributed by IGWMC from 1997 to 2005 and which was replaced by HYDRUS 2D/3D, version 1.x, in March 2006.

2/ Yesterday you sent me project Rio Bravo_2b.h3d (see above) that was created in HYDRUS 2D/3D version 1.x (note that the file extension *.h3d belongs to HYDRUS 2D/3D and you cannot create such files in old Hydrus-2D). I was surprised because I had expected a *.h2d file and I asked you about your license number for HYDRUS 2D/3D, version 1.x, that you had to use to create this project.

3/ Today you sent me again a figure of dialog A with text "I have attached the Hydrus info for our license. I am fairly sure we are using the old version of Hydrus". This does not make any sense because I'm sure that you created project Rio Bravo_2b.h3d in HYDRUS 2D/3D version 1.x. Since you did not send me your license number for HYDRUS 2D/3D (that you are using) you can hardly expect any support from us.

4/ If you explain me how you created project Rio Bravo_2b.h3d (using a legal HYDRUS license), I will help you to fix the geometry and FE-mesh in project Rio Bravo_2b.h3d. If you currently do not have a legal license for HYDRUS 1.x, I would recommend you to purchase HYDRUS 2.xx, level 2D-Standard.

Image

jsamson
Posts: 16
Joined: Thu Feb 02, 2012 7:14 pm
Location: USA

Post by jsamson » Fri Feb 10, 2012 5:44 pm

My issue has been solved. Thanks Mirek.

Mafer
Posts: 5
Joined: Sun Jul 22, 2012 3:59 pm

Post by Mafer » Sun Jul 22, 2012 4:11 pm

Hi. I have the version 2.008 of Hydrus-2D. I have the same problem of gloating point error. Mi domain is 4000 x 63 m. I need determing the flow in the domain for precipitation of 1 day. I tried to increase the spatial discretization, given that my temporal discretization seems like it matched up with the info contained in the link, but I still get the same error.

Jirka
Posts: 4712
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Post by Jirka » Sun Jul 22, 2012 8:56 pm

Considering how large your domain is, you will likely have problem with the spatial discretization. Please read the following text:

http://www.pc-progress.com/Documents/No ... zation.pdf

Jirka

Mafer
Posts: 5
Joined: Sun Jul 22, 2012 3:59 pm

Post by Mafer » Mon Jul 23, 2012 12:48 am

I tried to increase the spatial discretization, given that my temporal discretization seems like it matched up with the info contained in the link you shared. but i have the same problem. Are you help me??

Jirka
Posts: 4712
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Post by Jirka » Mon Jul 23, 2012 2:46 am

You should start with a smaller domain and figure out where your problem is there. J.

Mafer
Posts: 5
Joined: Sun Jul 22, 2012 3:59 pm

Post by Mafer » Mon Jul 23, 2012 4:02 pm

Mirek Thanks for your reply.

This happens to me only when I try to put a boundary condition at the top: atmospheric, variable flow, etc.

I never had happened, but never had a problem with a domain so long
and shallow.

I then make the model more profound? or increase the mesh in the upper?

Regards,

Jirka
Posts: 4712
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Post by Jirka » Mon Jul 23, 2012 4:24 pm

It all depends on the conditions in the surface layer and BCs. I have few questions, which I need to know, before I can answer:
1. What is your surface flux?
2. What are the soil hydraulic parameters?
3. What is the surface pressure head when the flux is applied?
4. What is average vertical discretization at the soil surface?

J.

Post Reply