Replies to This Discussion

Permalink Reply by John Brock on January 4, 2013 at 3:26pm

It looks like maybe your rails go vertical (parallel to Z-axis) or even past a bit. That would cause the sweep to flip.

JB

• ▶ Reply

Permalink Reply by Ryan Wynott on January 4, 2013 at 3:30pm

Yes, it does.

I tried to offset the path and do a sweep 2 but the offset went all garbage.

The website barfed on the attachment last time. Here's the file.

Needs to be roadlike so it will fit over a vertical surface trimmed to this shape.

Attachments:

• bad sweep1.3dm, 138 KB

• ▶ Reply

Permalink Reply by John Brock on January 4, 2013 at 3:41pm

Sweep1 needs to figure out which way is up with only 1 rail so it uses the CPlane normal. Going vertical is the problem. Try tilting your rail and profile a little to avoid the vertical section, do the Sweep, and tilt it back.

• ▶ Reply

Permalink Reply by John Brock on January 4, 2013 at 3:50pm

You can also add additional profile curves at the locations you want it to be a specific location, like on both sides of the area that flips. Maybe run the goofy sweep again, harvest a U/V isocurve on either side of the flip (before it goes wonky), delete the bum sweep, and sweep it again using the new profile curves you harvested.

Just an idea...

JB

• ▶ Reply

Permalink Reply by Ryan Wynott on January 4, 2013 at 4:10pm

Thanks John.

I tried tilting... and it just glitched in other places, although not the 180 reversal like previous.

I even gave up and decided just to use a pipe, and even that isn't working. Attached.

Something must be wrong with the curve. I just don't know how to repair it.

Attachments:

• grrrr2.jpg, 224 KB

• ▶ Reply

Permalink Reply by John Brock on January 4, 2013 at 7:19pm

Hi

I had a chance to poke around with your file and your long, wiggly curve is the problem. It has 70 sub segments. That's fine. The problem is the segments don't meet with tangency or curvature continuity. The ends are Joined but many of them are kinked. If you use the CurvatureGraph command and kick the scale up to about 125 or so, you'll see at the start sections where the profile curve it, it looks great. When you get into the corners the graph gets wiggly. It appears like maybe you did a lot of copying of the repeating sections so the wiggly problems were copied along. If you tun on the control points you'll also see how non-uniform they are. 

I tried FitCurve with a tolerance of 0.01 and it helped, but you'll see part of the curve gets wiggly.

Basically, you need to build a clean rail curve with as few control points as possible, that are uniformly arranged. It's a simple enough shape. Then I think  Sweep1 will work fine.

• ▶ Reply

Permalink Reply by Ryan Wynott on January 4, 2013 at 7:25pm

I got that curve by dupedging the other part.... then it went through some rebuilding. The other part was trimmed with curves from an autocad file. Not sure if that translates through with dupedge.

Can I salvage that curve in any way or do I have to start over?

Thanks for taking the time to look at this John.

• ▶ Reply

Permalink Reply by John Brock on January 4, 2013 at 7:50pm

Hard to say. If you turn on the control points look how they look in the first part of the rail on the profile curve end. This is the section that sweeps fine. When you get to the first sideways bend, look how goofy the points start looking and how many there are. That's where the problems start. SimplifyCrv did help some in my fooling around with it but I don't see any 'automagic' way to make that into a good, clean curve.

• ▶ Reply

Permalink Reply by Ryan Wynott on January 5, 2013 at 6:00pm

Thanks John. I ended up being able to rescue the curve by doing a blendcrv for each join. Kinda tedious but it worked.

Permalink Reply by John Brock on January 5, 2013 at 7:42pm

Excellent. If the blend curves pull too far from your original imported curves, you can probably use the EndBulge command to tweak the curve location without messing up the end condition continuities.