Shell not exact? All  1  2-19

 From: Michael Gibson 12 Sep 2012  (2 of 19)
 5385.2 In reply to 5385.1 Hi Rudl, you did not do anything wrong - that is a bug in shelling in the geometry library that MoI uses. The geometry library assumes that each vertex from the original object will match to one new vertex in the output object. But that kind of 1-to-1 mapping is not actually what should happen in these kinds of cases, it's something more like the pieces will diverge away from one another in that kind of situation and there should be things like new edges introduced into the result instead of a single vertex to single vertex output. So MoI's shell function is not able to properly deal with that situation, for something like that you may need to construct the offset by using Edit > Separate on the original shape to separate them out into individual surfaces, then offset those and figure out how you want to intersect them with each other. I have been asking the authors of the geometry library that MoI uses to update this area of the library for quite some time, and they have told me they have had someone working on it but that it is still in progress. - Michael

 From: Michael Gibson 12 Sep 2012  (3 of 19)
 5385.3 In reply to 5385.1 Also you may want to try using ViaCAD for some shelling cases like this that MoI's shelling is not able to handle - it might be able to handle it, it is at least worth a try: http://www.punchcad.com/p-9-viacad-2d3d-v8.aspx - Michael

 From: Michael Gibson 12 Sep 2012  (4 of 19)
 5385.4 In reply to 5385.1 Here's a bit to show you how this kind of shelling is kind of more complex than it first seems. Here is the result of doing a separate and then offset of your shape to the outside: When separating objects into individual surfaces like this each surface will be offset individually and there won't be any attempt to deal with extensions. But this way you can see how the sort of basic shape that each surface wants to have for its offset, and notice there how since each surface is at a different angle to one another that there is a different amount of space between the "natural offsets" ? That's the kind of thing that MoI's geometry library will not handle currently, it means that there will be a change in vertex/edge structure in those areas if they were extended and intersected with one another properly. - Michael Attachments:

 From: Rudl 12 Sep 2012  (5 of 19)
 5385.5 In reply to 5385.4 Also you may want to try using ViaCAD for some shelling cases like this that MoI's shelling is not able to handle ..................................................................... Thank you. But I don´t want to learn a new program. I will try it with TC, because it can import .3dm files.

 From: Rudl 12 Sep 2012  (6 of 19)
 5385.6 In reply to 5385.5 TC wasn´t able to shell it. I asked now in TC Forum, I hope the guys there can help.

 From: Rudl 12 Sep 2012  (7 of 19)
 I have been asking the authors of the geometry library that MoI uses to update this area of the library for quite some time, and they have told me they have had someone working on it but that it is still in progress. ............................ Hi Michael, you should press the guys. In my case it is a real showstopper.

 From: Michael Gibson 12 Sep 2012  (8 of 19)
 5385.8 In reply to 5385.7 Hi Rudl, let me know if after talking to the TurboCAD people it can't do it either and I will then show you how to do it by manual construction instead, which is pretty labor intensive but will be able to get the job done. That will involve some stuff like generating the intersection curves between 2 individual offset planes by using Construct > Curve > Isect, then doing some trimming of those intersection lines with one another and then using the resulting curves to form new faces built with Construct > Planar. - Michael

 From: Rudl 12 Sep 2012  (9 of 19)
 5385.9 In reply to 5385.8 Thank you Michael for your proposal. Yes, please show me, how to make it. An ex-and imported .sat file to TC was able to be shelled, but there is the same or a similar mistake. Is this mathematically not possible? Can´t believe it. In the attached picture you can see TC´s result. Image Attachments:

 From: blowlamp 12 Sep 2012  (10 of 19)
 5385.10 In reply to 5385.9 Can you post the SAT file here and I'll try to shell it in ViaCAD. Your previous file didn't import correctly for me to test it. Martin.

 From: Michael Gibson 12 Sep 2012  (11 of 19)
 5385.11 In reply to 5385.9 Hi Rudl, can you please post the model file of the result from TurboCAD ? Just looking at the screenshot you posted, that does not necessarily look like an error. It is a normal consequence of offsetting that a combination of planes at various different angles to one another will not produce a single vertex in the output while maintaining the same other boundaries of the existing surfaces - in order to produce a single vertex in the result it would require the planes to be extended some distance on their sides. This is due to the different sized gaps between the "natural offsets" of the surfaces, I was trying to show you that earlier in this thread with this message here: http://moi3d.com/forum/index.php?webtag=MOI&msg=5385.4 So if you want a single vertex in the result, you'll probably need to construct the extended region yourself, it will be easiest to do that if you have some good starting geometry to work off of which the TC result that you show a screenshot of may well be. Mathematically with constant thickness offsets being used, you can either have a result where each edge in the output tracks along the surface normal of the originating edge, which will produce a different edge structure in the result just like you see there in TC, _OR_ you can have the same vertex structure if the surfaces are extended by whatever amount necessary to their sides which means you will have a single vertex there but due to the extension the edges on the boundary areas will not track exactly along the surface normal from the original edges, those edges will be extended over to the side by some amount. It seems that maybe you want both of those at the same time, but yes as far as I understand that's not mathematically possible, you can get either same topology with extension/displacement of the final edges, or edges following along the surface normal of the original but with a deviation in topology as you show in TC. I guess maybe I would need to know more details about what you plan to do with the result of the object in order to be able to give you some tips on how to do it - do you need the "side walls" of the shell to be at a precise 90 degree angle to the originating surface? If so then the result from TC is what you want. Or do you want there to be the same vertex structure in the output but the "side wall" connecting the edge will flare out since there will be some material added to the offset in order to make the extended pieces connect up again in a single point? - Michael

 From: Michael Gibson 12 Sep 2012  (12 of 19)
 5385.12 In reply to 5385.9 Hi Rudl, or maybe another possibility is that you want some result where the offset is made more by expanding the original shape along some distance along the averaged direction from the original vertex point? That produces another different kind of shape in the result where there would be somewhat different thicknesses between every set of planes rather than a totally constant distance between every piece... CAD offsetting tools are more oriented around keeping the thickness between surfaces constant rather than trying to make a constant distance between the resulting corner vertices. So there are a few different variations in what kind of result you might be looking for, you will need to sort of nail down what you are looking to achieve a bit more specifically I think. If you want 90 degree side walls and a constant surface thickness, I believe the output from TurboCAD may be that result, that's the one where there will be some change in vertex structure in the offset piece. - Michael

 From: Michael Gibson 12 Sep 2012  (13 of 19)
 5385.13 In reply to 5385.9 You also might find that TurboCAD gives you the answer that you are expecting if you make your initial object be a totally closed solid rather than an open surface fragment. Right now from the result that you posted it seems that the thing you don't like is that it's not doing extensions on the border edges of your initial piece. If you have a closed solid there won't be any fixed-in-place border edges like you have with that open surface object and that will probably make TC extend all the different pieces in the result until they meet at the kind of structure that you want, if that is indeed what you want. If what you want instead is a 90 degree side wall between the pieces then like I wrote above the current TC result that you don't like is I think the right result for that case. - Michael

 From: Don (DON_CHEKE) 12 Sep 2012  (14 of 19)
 5385.14 In reply to 5385.9 Hi Rudl. I too have found SAT (and STEP) files exported from MOI work better than the 3DM file format when opened in TurboCAD. _________________________ Don Cheke Visit: Textual Creations

 From: Michael Gibson 12 Sep 2012  (15 of 19)
 5385.15 In reply to 5385.9 Hi Rudl, also probably if you had the original object a solid the offset would work ok in MoI too, since then the assumption of "one input vertex generates one output vertex" would hold true. It's in this case of a non-solid object with open boundary edges like you've got here where that "one input vertex generates one output vertex" assumption does not hold true anymore, and that's when MoI will produce a badly formed result, it will have just one vertex in there instead of it branching out into more than one with a new edge between them like you see in the TC result. So the actual goal for what I hope they'll be able to do to improve the offsetting is to actually generate I think the same result as what you don't like from TC... - Michael

 From: Rudl 12 Sep 2012  (16 of 19)
 Thank you all, @ Michael, I will tomorrow try to understand your explanations and will tomorrow also explain what my final goal is. The file Test2 shows in a simple way, what I want. The file sathaus3 is the file for testing with ViaCad THe file sathaus3 Shell is the shelled version with TC Attachments: