Thanks guys,

I have been trying various sweep stuff and I tried the rebuild command. I got something almost perfect with sweep. The rebuild didn't work out. It caused a lot of problems. Once I get this just right, I'll post how I did it and include some pics. I do think the upgrade to blend and network would be good. Allowing you to select multiple curves and hopefully do a g3 or higher surface. But the sweep method looks like it is going to work.

Here are some pics of a model that is really close. The only problem is there are problems with some surfaces that I'm trying to work out. All sorts of small slivers are being created in certain areas. I am using a mixture of sweep, network, rebuild, and other commands. Basically the sweep works but its a bit jagged in some places. So I'm trying to extract the curves and do a network. That is almost working. I'll keep working on it. I think it can be done it is just pretty difficult. Also I attached some pics of the small trailing edge radius issue. Right now I am increasing by about 6 times what is needed. The reduces the chord by about 1 percent.

I need to perfect the process to avoid the surface sliver elements but I can tell at this point the final model will look like the pics attached. The rebuild command is useful. I'm learning when and when not to use it. The multi-profile sweep is useful too. I wasn't familiar with it before.

Ideally however I am looking forward to an upgraded blend and network command.

Hi anthony, it's looking good! I was not able to see any of the slivery stuff that you were talking about in those particular screenshots. If it looks just a little triangular, it is possible that you're just seeing some display artifacts because the display mesh just to not have generated a lot of triangles in that particular area. That kind of display issue does not necessarily mean that there is a problem with the actual geometry.

The display system is generally oriented towards trying to do things fast so that you don't have to sit around for too long after doing every operation, so sometimes with this emphasis on speed of the display there is a chance for there to be some imperfections in it as well.

If it looks more messed up than just slightly jaggedy then it could be a problem though. It's hard for me to make a more accurate guess without seeing an example.

- Michael

hi michael,

they were so small i didn't get a screen shot of it. what is happening is the multi-profile sweep isn't giving me quite a smooth enough transition. It gets it exactly right but its rather linear in some places rather than a smooth curve. So I have been getting curves for a network patch by intersecting the results of the sweep with two planes. Then I use those curves and the network command. It looks much smoother but along a few lines it creates all these small surfaces. Too small to really show you. I think it is because of taking the kind of jagged sweep results and getting curves. It's hard for me to say when or why its happening exactly. What I am going to try next is rebuild the curves I need with a low amount of points and then run a spline through them. So make completely new curves. I think it will work. Everything is a band-aid though. It would really be great if the blend worked. I have to go through a lot more steps to basically do multiple network commands because the blend and sweep aren't quite working out.

A long time ago you were saying something about maybe making a tolerance that was adaptive or something. To get around the problem I have with the small trailing edge. I don't understand the issue even though you explained it to me a few times. But do you think there is something that could be created to allow for smaller features to be captured. Also this might play into the sweep issue I'm getting now. When I do the multi-profile sweep its almost perfect but it seems like I have a place where the rate of change is a little outside its comfort zone. So rather than a nice smooth curve I am getting like a piece wise linear deal. I'm envisioning perhaps sliders in dialogues where tolerancing is computed internally. maybe if it was exposed and the user can adjust it, that could possible fix many of the things i'm running into.

overall I think I can get this to work. but i'm just having to use a lot of trial and error and work around various problems that pop up. this has been good though making me get to know moi more. i really like it, even though its kicking my a** for three days now. i know a lot more about it than i did before.

Hi anthony, I'm afraid it's really difficult for me to understand what kind of straightness you were seeing without actually seeing an image.

From what you describe, it is entirely possible that the straightness was just a lack of triangulation in the display mesh in some areas which is just a display artifact to ignore and not necessarily an actual problem in the underlying geometry.

It's really hard for me to know for sure though just by a general written description.

- Michael

hi michael,

understood. here are some pics using the sweep. the back edge that has a ton of curvature appears smooth until you start zooming in and then it becomes piecewise linear. i have tried various mesh angles but this doesn't have an affect on it. the only way i can get rid of it is to go through a bunch of steps to generate curves that i can then use the network command on. it is much easier to use the sweep obviously. i tried rebuild on all the curves and rails prior to the sweep and that helped a lot. i previously hadn't thought to do rebuild on the rails. but i am still left with this piecewise linear problem. it seems like there is some sort of internal setting that the user can't access that is controlling this phenomenon. not sure though. in any event, i can see this working out ok. i just have to pound on it some more.

It might sound like a good idea to have a patch that can be G3 continuous at it's edges but looking at you prop design, it would be extremely difficult to use. Solidworks can do this but looking at your model you would need four patches. Now, which one would you do first? The first one is going to have two curves as two of the edges. These can only have G0 (positional ) continuity. Then when you put the next patch in, you face the same problem on at least one edge. The resulting surface may well not be what you want. These type of patches (certainly in my experience with SW) are normally used to close a hole that has been delberately cut out to remove unwanted geometry.

Hi anthony, so from what I see there that does look to be likely just a lack of refinement in the display mesh in those areas.

It's most likely nothing to be concerned about - in order to be displayed the surfaces have to be converted to triangles and those areas of the surface just did not get a whole lot of triangles generated for them.

The display is inherently piecewise linear because video cards only understand how to process triangles. Most of the time MoI does a nice job of making enough triangles to make things have a smooth appearance. In some situations you may not get enough - that's partly because the focus on the display mesh generation is on doing it really quickly so it cuts corners in some areas and it is possible to have a bit of a jagged appearance when that happens.

When that happens it is not something that you need to go through any great lengths to fix - it's not an actual error in your geometry so you should just ignore it and not try to reconstruct your model.

Normally one way to verify that it's only a rough display mesh would be to export to a mesh format and crank the export density way up and look at what that does. The export mesher uses a somewhat different approach than the display mesher and does a much more thorough job of testing and refining things.

But my main advice for what you see there is to just ignore it - as far as I can see it's simply a display artifact and not any actual error. Maybe MoI's typically nice display has inadvertently made things worse in this case, by making you worry too much about seeing some evidence of the slightly lower density triangulation in the display mesh in some spots. I mean what you are showing there seems to be a close zoom in to a small area of the large surface, just doing that kind of closely zoomed in examination will exaggerate what is actually only really a small amount of angularity in the display...

- Michael

The curve display is somewhat different though - curves are diced up dynamically to only deviate by less than a pixel on the current view screen, so the curve display automatically refines as you zoom in.

So that means that what you see on your screen for curves is a really accurate representation of the real curve.

Unfortunately there is too much number crunching work involved in doing that same dynamic refinement process on shaded surfaces, so a surfaces are only broken down to a static set of triangles one time and do not automatically refine as you zoom in. So that's why it is not unusual or an "error" to see some evidence of the triangulation process in the shaded surface display. When you see some slight angularity and it does not look like it is strangely leaking outside of bounds or looks like really scrambled colors (those types of things can be evidence of trim curve problems or self intersections), then you're just seeing some of the silhouettes that the display mesh is made up of. That's nothing to worry about - it's a normal part of the display.

- Michael

thanks guys,

you may be right steve. i have never tried this before so i don't know how to actually do it. it seemed to me that a blend with higher continuity would make sense. however the sweep is doing what i want. i'm not sure how it figures it all out but it is doing it.

i do get a lot of graphics issues with the surfaces. so its good to know that it is just graphics and not a real problem. it may be nice to be able to see the actual surface mesh in moi though based on what you are saying. at least to be able to check the real thing without having to go to another program. however, i've got what i need at this point. again my only purpose was to kick the tires on moi to see if its a good downstream program for PROP_DESIGN. i believe that it is. just a little tricky in areas.

Hi anthony,

> it may be nice to be able to see the actual surface mesh
> in moi though based on what you are saying.

You mean the surface control points? You can do that in MoI by separating out the surface so that it's not joined to anything else and then you can turn on the control point cage with Edit > Show pts.

But that shows you the control polygon hull of the surface - the surface itself is a smooth spline that is influenced by those hull points.

Your video card does not know how to show a mathematically continuous spline directly - any program that you load the surface into is going to need to break it down into triangles in order to display it just like MoI does, and most of the time they do it actually quite a bit rougher than MoI does. That kind of slightly jagged angular type silhouette is a pervasive type of display artifact that's common to pretty much all CAD programs.

It's just not something to really worry about so much - as long as the only thing that looks weird is a slight angularity but things are not leaking grossly outside of boundaries or look like pieces are sticking through each other then it's pretty safe to assume that you're just seeing a slightly coarse display mesh in that particular area.


> at least to be able to check the real thing without having to
> go to another program

Normally the best way to check it in MoI is to export to a polygon mesh format like OBJ for example and then crank the export mesh density way up. The export mesher does not take shortcuts with how it analyzes the mesh and decides to subdivide it, so you can use it to generate a more accurate polygon representation and examine that. But you won't be able to use that method with the "no save" version though, since it relies on going through the process of saving to a poly format.

- Michael

thanks michael,

i posted the final example pics on my website:

http://propdesign.weebly.com/screenshots.html

i mention MoI on my website as well at various places. All good words of course. This does it for me. Fantastic program. Thanks for all the help. I couldn't have done it without you. I couldn't even import the points at first. So the forum support was definitely important.

No problem anthony, I'm glad that you are liking MoI!

- Michael

hi michael,

hard not to love moi. here are some pics of the surface definitions from moi using the instructions you gave. awesome feature. this really explains the piecewise linear thing to me now.

Hi Michael,

I played around with MoI today and got an even better result. Sweep, fillet, revolve, and boolean ops were used to create what you see. It was very easy and fast to create, once you know what to do. Basically (as you stated earlier) the fix for this thread topic was instead of thinking of this as a blend op, think of it as a sweep op. Then everything is great.

Thanks for the help. Awesome program.

Anthony

Anthony,
Your program is really going to turn out NICE!!!! I always like seeing your threads come by.

Thanks burrman,

Your help has been very useful. PROP_DESIGN is essentially complete. I am only doing minor maintenance to it these days. Recently I recompiled it to get a little more speed.

I have been learning MoI some more and am able to make better models. I'm still learning though. I tried things a slightly different way and got even better results. The only problem I'm having right now is figuring out what causes bad surfaces or bad visualizations to occur. It seems rather random. Some of the graphics problems I reported earlier in this thread I can make go away depending how the model is made. But I don't have a handle on it yet. Some times rebuild helps and some times it makes things worse. The way that you go about the sweep changes things a lot too.

I'm going to mess with it more until I get something repeatable. Here are some pics of a more accurate model. The trailing edge fillet blends into the shank much more realistically now. Out at the tip of the blade the model is all distorted, when you zoom in. So I have to figure it out more. I didn't show that in the pics.

I have all the latest screenshots on my website.

Hi anthony,

> The only problem I'm having right now is figuring out what
> causes bad surfaces or bad visualizations to occur.

But keep in mind that the problem of a little bit of polygonal looking silhouette in the visualization is a normal part of the realtime viewport display mechanism and is not something that you should actively work on avoiding.

- Michael

Hi Michael,

I believe my problem is stemming from the following, based on my recent tests. It seems I have to increase the trailing edge radius (ter) by a factor of 3 and break the circle up into four pieces. Doing these two things allows the sweep op to create the desired geometry. If I don't scale the ter at all then I get very random results. If I scale the ter by 2.5, MoI will create geometry in a repeatable fashion. However, not the geometry it does when the ter is scaled up by 3.

Once I got a handle on everything, all that you have to do is select the circle, select the airfoil and do a two rail sweep. If the ter is big enough and the circle is broke up into four pieces (presumably to match the four airfoil curve segments) then everything is accurate, repeatable, and displays properly. Rebuild and mesh size had some affect, but not really anything related to my problems.

On a different note:

I realized that having the variable pitch blade shank on the swept blade does not make sense in reality. I was doing it just to kick the tires on MoI. But for the sake of reducing confusion I put my novel fixed pitch blade attachment on the swept blade and the traditional variable pitch attachment on the straight blades. The swept design is for the Airbus A400M military transport and the straight blades are for the Piaggio Avanti II business turboprop. Both very cool planes.

Thanks to everyone for your help, I believe I have everything under control now. I'm going to remove all the old pics I attached to save server space. As always these pics are also on the PROP_DESIGN website.

And just to summarize this thread; basically the answer was not to use blend, use sweep instead.

I created a screencast showing how I made the Airbus A400M model. I plan on adding one for the Piaggio Avanti II as well. The screencast can be found here:

http://www.youtube.com/user/propdesignsupport

It's about 15 minute long and there is no sound. In this case I scaled the ter by 2.5 and rebuilt the ter and the rails. This prevented downstream errors that would occur. Scaling the ter alone did not fix the problem. The rebuild was necessary.