I was rebuilding a complex closed polysrf I had build in V7 in V8. In the rebuild, I am not getting a closed polysrf. The openings are points.
I exported to V7 the surfaces join without creating point openings.
The old TestRemoveAllNakedMicroloops (RemoveAllNakedMicroEdges) used to fix that kind of thing but I am getting no effect.
Is there some way in V8 to close this point holes?
these surfaces are frankly really bad… you need to do some rebuilding to clean these up to have any hope of getting a decent result from something like this.
what is the goal? are you 3d printing these? if so, abandon trying to clean this up and simply shrinkwrap it to get a closed mesh you can print.
if you absolutely have to get this closed up in nurbs, an ugly but effective cheat is to place a cube on that point, use it to trim a hole and then close up the hole with a 4sided surface of your choice (network, edgesrf, 2 rail sweep, etc…) to be clear I do not recommend this, but desperate times call for desperate measures, and I’ve been there done that.
This is being used for a lot of different things. I was previously used for docking the the ship and locating items. It may be used to deterine where to cut openings for visitor access. It is also used for illustrations showing how parts fit together.
3D printing might be on the horizon.
Here is the problem I encounter in ship hulls regarding complexity:
In places of tangency things come out pretty simple. The vertical isocurves shown are about 4-feet apart density. I can do this with sweep2. Unfortunately, warships tend not to have a lot of areas with tangency.
With a concave curve, I can get away with about 9 points and be reasonably accurate. Things get denser but the vertical curves shown are about 1-foot apart. With these, sweep2 starts to create accorian pleats and I have to switch to networksrf (for all those who say never use networksrf). (The horizontal lines at the extreme left of the image are knuckles, requiring separate surfaces.)
For convex curves at least 13 points are needed (possibly even more). The vertical curves shown are still about 1-food apart. Accordian pleats are really bad here as well with sweep2.
While curves appear dense, over a 200+ -foot long segment, the curve density is rather rather low compared to size. A 1/16" variation from ideal is visible. This is a 3 1/2-foot by 2 1/4-foot enlargement
Large segments may be simple and fun. But when you need to get around propeller shafts and other tight areas, the density explodes (as in the opening example). Many small surfaces are required to create the shapes in tight corners. Sometimes it is impossible to avoid 3-sided surface. One can get away with edgesrf irregularly but most of the time one needs networksrf to get the basic shape close before using matchsrf. Here, the curve density shown gets down to about 1/2". Each surface filling a patch grows the density. Each Matchsrf with refine match grows the density further.
Is there some technique for keeping the complexity from growing in these patches around tight corners?
I am running into this problem because I was adding some new internal structures and discovered the hull was 1/2" too narrow at the bow due to incorrect tapering. I had to correct the frames to more closely fit the plans. Then I had to rebuild. That is where I discovered that which was building fine (in spite of complexity in tight corners) on V7 was not building on V8.
are you reverse engineering a scan? I’m trying to figure out why your curves that you use to generate the surfaces are so dense. Most of those patches seem to be able to be described with a deg3 4 pt single span curve… I’m sure I’m missing something here…
Especially on something so big, you have a lot of tolerance to play with.
Rhino has lots of bugs regarding surfaces that are supposed to come to a point.
Nobody at McNeel cares.
I suspect that if you go to a parallel view and zoom way in you will be able to see what is the problem.
post the surfaces around the point failure if you want someone else to take a look.
Based on his previous posts @miano is starting with the original detailed US Navy plans of the USS New Jersey battleship . He is working to the plans as accurately as possible create a digital representation of the ship as built, including each component of the structure. The plans have much more informaton that what is needed to create a basic model of the shapes. My understanding is he is attempting to keep the digital model as close to the plans as possible.
This is not a simple task, and is very different than creating a new design, or even creating a digital model of an existing object. such as a car, based on photos, observations and possibly simple drawings (frequently from publications or advertising brochures and of dubious fidelity) with the intention that the digital model “look the same” as the original object rather than match the original within a tolerance.
thanks for clarifying-
To add, the amount of tolerance that I can get away with depends up the area. Even the real thing was off a few inches. However, when two thing need to fit together the tolerance gets much more imporant. The model was used as the basis for ultrasound testing of shell plate thicknesses last spring when the ship was out of the water to see if which places had severly eroded. In that case, the measurements are fractions of inches. Fortunately, there were only a few such spots that were corrected by welding new plating over the weak point.
In any event, the large expanses are doable with relatively sparse modeling. When it comes to the tight corners (the parts no one shows in ship videos), networksrf using an edge, matchsrf/refine, arcs as edges, start exploding in complexity. I start with edgesrf but there is enormous trial and error because the various patches do not fit right or I get a bad surface when joining. EdgeSrf in a curved area tends to create a surface that bends against the desired curve.
Much of the problem arises from needing a curved surface where the radius goes to zero. These create dreaded 3-sided surfaces that have a tendency to cause other problems. I wish there were a good rhino solution for these.
I exported the file to V7 and RemoveAllNakedMicroLoop appeared to fill the holes that it was not filling on V8.
I have torn this think apart to try a little cleanup but I do not have a good side-by-side comparison at the moment.
3sided surface areas have a number of pretty common solutions, please post one and I’ll be happy to try and help show the fix-
If I can’t solve it, @skyg likely can .
This is the three-sided alternative that use most frequently:
The 3-sided problem I have never solved is this: radius going to infinity at a corner:
can you post those surfaces?
Problem Triangle.3dm.zip (3.3 MB)
in this case sweep 2 run on rebuilt surfaces yields a good result IMO-
I’m not sure if this deviation is within your tolerance, but I guess I keep going back to the fact that these surfaces are massively over defined and are the root of all your modeling issues. IMO you are working way too hard for the sake of accuracy, where in fact you are not actually getting what you want because your surfaces are too dense.
Maybe I’m confused, but this seems like a classic reverse engineering job where you have to balance accuracy with quality. Every RE job I’ve ever done (and I’ve done many many of them) the first step is to clean the data set so that you can make reasonable surfaces and get a decent result.
but, there is likely much about this job I am not understanding, so take my thoughts with a grain of salt.
