Michael, thank you so much for taking the time to explain that to me. I would not have thought to model like that had you not.

However, one step I do not understand though is where you Construct / Offset / Shell and then use Construct / Boolean / Difference to give the model volume. I don't get how you've done this? The offest seems brilliant but it doesn't 'close' the model or give you the option to? Also using Boolean Difference I don't see how you've managed to close the object?

I hope you can spare me some more time to fill in these last parts of the puzzle for me!

EDIT! Hold everything, just seen the 'Shell' submenu in Offset! What an idiot.

Anyway thats sorted now. I've got my shell looking great with thickness and a fillet. Now I've sweeped a circle along the shells outside curve and then attempted to Boolean/Union the 2 objects. Unfortunately it doesn't work? I've applied ajoin to both object but it hasn't made any difference. Furthermore I can't seem to 'jion' the sweeped circle? It only stays in 2 halves?ve attached the current file.

Hi gizmo,

> Furthermore I can't seem to 'jion' the sweeped circle? It only stays in 2 halves?ve

This is how I see it, the curve that is extracted from the model is quite complex as in, if you turn on the control points for this curve you will see there are a lot of points, so I think the auto solver for the sweep is having a hard time calculating a nice sweep, and if you zoom up into the joins, circled in red, you will see that the ends have gaps and are not square or plumb.

To get this sweeping more accurately, I again only worked on a quarter of the curve and placed the circle profile on each end of the curve 90 deg to each other, then swept it with no caps, then mirrored and joined.



I've attached your file with a completed sweep and one set up ready for sweeping on the model so you can see what I mean.

~Danny~

Thanks DannyT. Right I see, I didn't realise that the rail I extracted would be so messy nor that its messyness would have an impact on my sweep operation.

Danny how did you break up the rail into 4? I don't know how to do this?

Is there anyway to cutdown on the messyness of extracted curves? I just took it for granted that the curve I extracted (via selecting the shells curve and copy and pasting) would be okay? I get the feeling that once an operation is performed on a surface it automatically creates masses of points? This is certainly something I wasn't expecting with Nurbs. For some reason I thought Nurbs would be immune from this type of thing...

Hi gizmo,

> how did you break up the rail into 4? I don't know how to do this?

Here's a quick video.

Cheers

Thanks for that Danny.

Right I'm still struggling with the peculiarities of this stuff. I've created a loop and I want to sweep a cirle around it and then union jointhe result from the shell.

Unfortunately when I try sweeping the circle it creates 2 half sections!? Why is this? I don't understand it? It creates a tiny gap between the 2 and it seems impossable to join them up again, using the join command doesn't work? Is there no manual weld or something like that?

Beause it making 2 sections when I come to union join the tube with the shell its going nuts and even more so when I try to fillet it after!

Could someone explain to me what's going wrong? I just don't understand why it seems so picky!?

Hi gizmo,

> Could someone explain to me what's going wrong?

To get a smooth sweep juncture, it is not enough to just have 2 points touching one another. To give you a bit of an exaggerated example, here are 2 curves that touch each other at a common point just like points you have labeled in red above:



Even though these curves touch at a common point, if you sweep them you will get this result:



You can see more easily in this case why the pieces do not mate up even though the 2 curves connect at a common point. It is because the curves are not totally smooth with one another.

You are running into this exact same issue with your curves - your pieces do not share a common tangent similar to the above, although your one happens to be much closer to being smooth so it is harder to see it initially.

A curve's tangent direction is defined by the end point and the next point to the inside.

If we look at these 2 points:



And then these 2 points:



Creating 2 extended lines that go through each of these points (here I used construction lines to do this):



Initially that looks pretty much like they are aligned, but if you zoom in:



You can see there is actually a deviation there - those curves do not have a common tangent so they are causing the same kind of effect as in the 2 sharp lines that I showed above.

I've attached a version of this curve pad_trued_tangents.3dm, where I lined up all the points to be exactly along a straight line (I drew in an extended line and snapped them all to it), with this version you should now be able to make a sweep where all the pieces join up exactly.

- Michael

Once again, thank you Michael for explaining that too me.

I'm beginning to feel that things like that really need to be made either more clearer or made so they can't happen? Perhaps adding a command which will automatically 'smooth' 2 selected points? I know that Align is available but imo this sort of problem shouldn't even come up. Again, I never thought I'd need to be as precise as this using Nurbs.

Ok so I've got my rail sweeped and the shell ready to be Boolean/Unioned together, which I do. I then select the top inside curve to apply a fillet to and... nothing. I can't get the fillet to work? I get the feeling that something isn't lined up correctly or there's a tiny part of the curve I'm unable to select? If so how am I able to get around this?
Here's what I'd like to fillet and I've included to file too.

Just in case people are losing heart in this thread and maybe feel its not worth replying to anymore.. please don't! I'm really trying to understand this and I promise to write up a explanation when all my problems with this are solved. After this I feel I'll have a lot to say that might make future versions better and also help other people understand Nurbs quicker.

Hi gizmo,

Unfortunately Nurbs modelling does have to be precise but I've noticed more so in Moi, I had no problem filleting your model in another 3d app but the cost difference is about 50 times.

Where there are seams involved, I've noticed in Moi if they are not lined up exactly it has a hard time solving fillets, so as you can see if you zoom up to the area I've circled the seam of the swept piece is out a bee's prick to the main body seam, so the way you correct that is to line up the 2 points from the top view because they weren't exactly aligned and then the fillet will work up to 6 units.

You get used to the accuracy required once you've understood nurbs modelling, at work we model to an accuracy of 0.005mm which I think is too fine but I have to follow protocol.

I've attached your model corrected for reference.

Oh yeah, you can fillet the right side of the model you supplied because the seams are aligned but not the left side because of the misalignment as explained.

Michael, any comment on why we have to be so accurate in Moi and how other 3d solids/nurb apps are more forgiving or stronger in solving such issues, is the Solids++ kernel still in it's infancy, will it get better as time goes on ?



~Danny~

Thanks for that Danny.

Could you explain how you aligned that seam? I can't see how you can snap the rail point to the shells point?

Hi gizmo,

I just eye balled the alignment of the sweep from the top view then snapped the seams in line, it was just quick for the exercise, I didn't really know where you wanted the sweep exactly as long as I got the seams to line up.

Oh yeah, another point I wanted to make about the seams is that it would have filleted if the seams were further away as well, like if it was 5 or 6 units away from the main body seam, it would work also, it's just that when it is so close it gives MoI a brain strain ;) Michael might correct me on this.

Cheers