topic Re: Mesh volume calculation accuracy? in Modeling

Mesh volume calculation accuracy?

R Muller — Wed, 24 May 2023 18:00:22 GMT

I am having inconsistent results in calculating volumes of cut and fill, as compared to the "2D" way of doing it by measuring areas outlined by "before" and "after" contour lines. I am beginning to suspect that the mesh tool itself may be inaccurate, and that I shouldn't try to rely on SEO between existing and new meshes to give me sufficiently accurate quantities of cut/ fill.

As an experiment, I created a mesh with fairly straight contours, defined by just a few points. When I added additional points to the existing contours (no change in elevation, just put another point on the existing line) the volume of the mesh changed. It changed even more when I added a "valley" line that cross-connects the contour lines along the valley that I am creating with my grading.

Is anyone else having an issue with the accuracy of cut/ fill calculations based on meshes?

Re: Mesh volume calculation accuracy?

David Maudlin — Wed, 10 Feb 2010 15:28:50 GMT

R Muller:

You might try changing the mesh 2D display of Ridges from Show User Defined Ridges to Show All Ridges. When you change this setting, you will see that ArchiCAD is dividing the Mesh into triangles connecting all the points, and as points are added and deleted, the division of the mesh changes, which would impact the volume calculation. In the image below, I have (2) 4' x 4' x 2' meshes, with three of the corners at 0' elevation and the fourth corner at -1' elevation. Depending on how the mesh is triangulated, the volume is going to vary.

David

Re: Mesh volume calculation accuracy?

R Muller — Thu, 11 Feb 2010 15:35:52 GMT

Yes, David, I have noticed that ArchiCAD seems to prefer the interpretation in your left example to the one on the right, which is why I have learned I need to add ridge and valley lines in these situations. Do you suppose there is a preference for flat ridges over sloped ones?

What I am wondering is if there is any sort of systematic bias, that would tend to show less cut and more fill, for example.

I am getting different results in analyzing a complex grading plan than my consultant, who is using a 2D method. I suspect my use of ArchiCAD mesh volumes may be incorrect, but I can't see where.

One problem is that the calculation involves subtracting two large volumes to calculate a much smaller amount of cut or fill, so a 1% systematic error in the volume calculation would be magnified many times in the cut/ fill calculation.

Re: Mesh volume calculation accuracy?

Anonymous — Thu, 11 Feb 2010 15:48:46 GMT

Cut and fill calculations by any means will always be an approximation depending on the accuracy of the survey and the assumptions of the process. To get more accuracy you need to add more detail to the mesh. If you lack the survey data to support this it will still be a best guess (which is what contour lines are anyway).

Re: Mesh volume calculation accuracy?

R Muller — Thu, 11 Feb 2010 16:30:18 GMT

I am trying to understand some of the parameters that lead to greater accuracy. I am modeling a 10+acre hillside site, with vertical elevations of more than 70', and I have a survey with 1' contour lines. If I put all that into the model everything slows to a crawl, and I can't show anything live in 3D.

So, I'm trying to strike a reasonable balance between accuracy and speed. For example, I believe that a uniform slope doesn't need every contour line, but the top and bottom of each slope should be put in, even if they are not at regular contour intervals. Ridge and valley lines need to be added, even though they are not contours.

Does anyone have some other rules of thumb?

Re: Mesh volume calculation accuracy?

Karl Ottenstein — Thu, 11 Feb 2010 17:12:09 GMT

R wrote:
I am trying to understand some of the parameters that lead to greater accuracy. I am modeling a 10+acre hillside site, with vertical elevations of more than 70', and I have a survey with 1' contour lines. If I put all that into the model everything slows to a crawl, and I can't show anything live in 3D.

I have no problems with much larger and much steeper sites - going back to version 7.

It is not necessarily the 1' contours as it is how you are creating the contour lines. If you are magic-wanding a curve with a very small tolerance, you will get an insane number of polygons. Depending on the source survey, you can use the magic wand with a looser tolerance, or if you cannot magic wand because of line breaks, then trace the contours manually without creating an insane number of points at curves.

Use the free PolyCount goodie to read off the number of polygons in your current mesh.

Cheers,
Karl

Re: Mesh volume calculation accuracy?

Erika Epstein — Thu, 11 Feb 2010 17:12:32 GMT

Are you developing the entire site? Or just parts of it?
One strategy is to upgrade your computer
On a more serious note,
-you can divide up your site mesh into sections
- you can model only those areas of the site that need 1' contour detail level and the rest at 5' or 10' intervals as appropriate, or some combination there of.

Also, keep your site and building(s) in separate files hot-linking one into the other. This is fairly standard as it allows you to orient you building to Project North on you screen and layouts, yet have the building sit oriented correctly on the site per real North.

Or a combination of the above.

Re: Mesh volume calculation accuracy?

R Muller — Fri, 12 Feb 2010 02:21:00 GMT

I have about 22 meshes in my model (for 5 different grading options), and each mesh has about 4000 to 4200 polygons. Typically only 2 meshes are turned on at once. but there are various SEO's with other meshes and slabs. I don't think this is a lot, but maybe you think otherwise. I do this lot, so I know the issues about magic-wanding the survey polygons.

I am using 10' contour intervals in steep parts of the site that I am not developing, and 2' intervals elsewhere.

There are no buildings, but I do have a lot of (simplified) trees. You can see some of the images on my client's website: http://ebmud.com/sites/default/files/pdfs/Summit%20Design%20Alternatives%203D_0.pdf
And some simple openGL movies at http://ebmud.com/project-updates/video-illustrations-summit.

I do this kind of site design a lot, but this is the first project where I have used ArchiCAD to directly calculate volumes of cut and fill. Much to my chagrin my results seem to be not valid. I say that cut and fill balance, while my consultant says there is twice as much cut as fill. We are not even close in our quantities. This is embarrassing!

Re: Mesh volume calculation accuracy?

R Muller — Fri, 12 Feb 2010 02:33:31 GMT

This link may work better for the movies:
http://ebmud.com/about-ebmud/news/project-updates/summit-reservoir-replacement-project
Click on the "video" link at the bottom of the page.

Re: Mesh volume calculation accuracy?

Anonymous — Fri, 12 Feb 2010 14:48:11 GMT

R wrote:
I do this kind of site design a lot, but this is the first project where I have used ArchiCAD to directly calculate volumes of cut and fill. Much to my chagrin my results seem to be not valid. I say that cut and fill balance, while my consultant says there is twice as much cut as fill. We are not even close in our quantities. This is embarrassing!

This does not sound like a result of rounding errors or the accuracy/detail of the site model but more like either a procedural issue or a bug in the calculations. You should be able to check this by scheduling the volumes of the various site meshes and checking them against some thumbnail arithmetic.

Re: Mesh volume calculation accuracy?

David Maudlin — Sun, 14 Feb 2010 13:17:19 GMT

R wrote:
Yes, David, I have noticed that ArchiCAD seems to prefer the interpretation in your left example to the one on the right, which is why I have learned I need to add ridge and valley lines in these situations. Do you suppose there is a preference for flat ridges over sloped ones?

What I am wondering is if there is any sort of systematic bias, that would tend to show less cut and more fill, for example.

Rather than adding lines to the mesh, I have added points to break the triangles into more realistic grading shapes. I don't have any feel for how ArchiCAD triangulates the site, but I have seen many instances where the triangulation yields elongated triangles (2 very long sides and 1 short side) rather than triangles with 3 roughly equal sides. I add points to break the elongated triangles into more even shapes.

[As an example of lack of control, when I created the meshes in my earlier post, I created the first one and set the heights, then copied it while wondering how I was going to force it to triangulate on the opposite axis. The copy divided the other way on its own, go figure.]

I have been creating meshes for visualization only, not for volume calculations, so I don't know about the methods used for cut & fill balance. I think you will need to understand the calculation method used by your consultant and be able to relate that to the ways the mesh is formed in ArchiCAD to see a way of editing the mesh so the two methods are closer.

HTH

David

Re: Mesh volume calculation accuracy?

Karl Ottenstein — Sun, 14 Feb 2010 20:59:55 GMT

R Muller,

Interesting project and issue. I have never modeled a terrain mesh as a series of plateaus as you have done, but as a smooth (fit to user ridges) mass that attempts to interpolate to reality. By stepping your terrain, the value error could be significant.

The only other thing I have to wonder about is the fact that you have 22 meshes and are doing SEOps on them. Typically, I'll have two - the complete 'before' mesh (smooth) and complete 'after'.

Nice to see planning work like yours being done in AC. 🙂

Karl