Showing results for 
Search instead for 
Did you mean: 
Showing results for 
Search instead for 
Did you mean: 

Inaccuracy with Spine Bend


Inaccuracy with Spine Bend

This week I tried something that I've been anxious to get to for a long time.

I built my 1st Spine Bend model.

At 1st I was elated, but then I compared the model against the original model which has already proved out with accuracy.

The bend is good with simple rotations but the angles off the "V" shape weave inside and out of the correct position. (The largest error is .075, we need to hold within .010)

See picture of the trouble angles.

Spine Bend.JPG

As far as I can tell the angle itself looks fairly good, it's just appears that the "V" position weaves in and out.

There isn't much documentation of what the Options do, perhaps one of these Options might help.

I have watched Leo Greene's excellent You Tube tutorial, but it doesn't hit on what I'm needing to control.

Would anyone know how to stabilize the "V" position on a Spine Bend?


Re: Inaccuracy with Spine Bend

You might be seeing an artifact of stretching the material. What if you reverse the spine so it runs on the outside or centered?

For a simple cylindrical bend, you can compare it to a warp feature.

Re: Inaccuracy with Spine Bend

You might also try a curve analysis. Change the part accuracy and you will likely see better results with the analysis as you tighten up the accuracy.

Re: Inaccuracy with Spine Bend

I don't believe I'm seeing an issue in regards to the model not being clean. The integrity of this bend looks better than expected. Also, my part accuracy is set more fine than the default value, it is .0005.

Honestly I expected to see variation in the process because of what is being done. How can you maintain rotational and positional integrity on faces that were angled prior to the bend? The mathematics would have to be very challenging by just having a spine curve to bend features that are angled at a tilt.

On the other hand I want to believe there must be some way to add influence to these angled faces to keep them positionally correct.

In our ZW software we are able to get the true angle and position of the angled "V" by using a completely different process. In ZW we build our geometry already rotated. In the early days we discovered that the angled lofts even though they laid perfectly on the drive curve the angles distorted along the drive curve loft. Through the assistance of technical support we discovered that by adding spine curve influence that we could get perfect rotationally integrity on the rotated angled faces.

I know I am doing a completely different process but I'm hoping it is possible to reign in the accuracy.

I'll attach a couple snapshots that show the accuracy deviations.

You can see how the drive angles pull in and out below. I have the original ZW geometry in magenta and the creo geometry in blue.


Below is a measurement edge to edge of the difference I'm seeing.

Error Distance.JPG

Re: Inaccuracy with Spine Bend

Paul, did you get this solved? Can you post your model for me to get a reference?

Maybe spinal bend isn't the answer. However, noting deviations, or causes, would be a good exercise.

Re: Inaccuracy with Spine Bend


Thank you for reminding me of my post.

This ended up much different than I expected.

All these years we determined that the geometry that we built using the "long" process in ZW CAD was accurate. We couldn't see why generating the 2 curves to the "V" on the rotating face with a point to point curve would be inaccurate. Yesterday I went through the process of generating equal distance points along the "V". To my astonishment I saw variation up to .125.

Having this surprise fresh on my mind I made the same test to the Spine Bend built in Creo. I was astonished to see that the positional accuracy was exact up to 4 decimal places.

Seeing that this number was near perfect I made the same test to the top intersection of the angle off the "V". Once again I was amazed to see the accuracy was near perfect.

I went into this experiment expecting to see deviation in the Spine Bend just because of the math involved in taking a straight solid and converting it to a roll diameter around a curve. If the deviation was small enough I would still decide to go with the Spine Bend method, because it was much simpler than building everything on a rotation. When the deviation was greater than what would be allowable I was about ready to throw out the Spine Bend process because it was inaccurate as we all expected. I submitted this case to PTC. One of the questions he had was "How do you know the ZW geometry was accurate". Because of the need for getting other work done I almost bypassed the testing, but I'm glad that I didn't.

The down side is that years of tooling have been released with a method that wasn't exact. The variation was small for rolls so big, but these deviations have the possibility of relating to production defects.

I am very excited with the prospect of changing all of our new work to Creo's Spine Bend. The joy is tempered a little bit however with the possibility that not everything we do with Spine Bend might not convert accurately. Only time will tell.

Re: Inaccuracy with Spine Bend

Wow, now THAT'S an interesting development! Being a big fan of spinal bends, this one intrigued me, I just didn't have time to play with it. Glad to see it was your other software that was the issue, hope you didn't make too much scrap!

There ARE other ways to do it. You could have wrapped a v-shaped curve around the cylinder, and used that as a trajectory for a VSS surface, with the angle related to an axis or the surface, and the section normal to the spine trajectory or something else.

Re: Inaccuracy with Spine Bend


Yes, I can recall another post I've seen that you have done some interesting work with Spinal Bends.

Prior to getting Creo we heard that others have used Spine Bend in almost an identical application as what we needed. Our group was convinced prior to buying Creo that it would be impossible to do an accurate flat to circle bend with the detail we needed, but it was intriguing just the same. We even convinced the technical application salesman that this would be impossible.

The beauty of doing what it appears is possible is that our layout we use to build the geometry is flat. We had to translate all of this to a round roll, this was a pain.

If the Spinal Bend didn't work we would have had to resort to something like what you suggested in wrapping a v-shaped curve around a cylinder for development.

Re: Inaccuracy with Spine Bend

Ditto what Frank said. I did some work in the past on barrel cams. From the testing I did the spinal bend wasn't nearly as accurate as the variable section sweep (coupled with a graph and trajpar). Of course the "twisting" that the spinal bend had to do to create the cam slot was much more complex than this seems to be.

Re: Inaccuracy with Spine Bend


This is good to know, just in case this doesn't work out as we still might find.

I am wondering if it is easier to achieve accuracy to a perfectly round roll?

Re: Inaccuracy with Spine Bend

Paul, thanks for the update. This is good news, at least going forward.

One other option... Make two cylinder surfaces and wrap your cam profile sketch onto each. Join with boundary blends and solidify.

This may be easier if you have some complex profiles. This also works for VSS by providing the additional "guides".

Always remember to include a CSYS -IN- the sketch to be wrapped.

Also remember that wraps work on non-cylindrical faces.

Re: Inaccuracy with Spine Bend

I'll certainly check back with you guys if the method I'm using has a failure point.

I'm showing the simple beginning of the model. There are more complexities than what are shown.

Re: Inaccuracy with Spine Bend

One more thing I forgot to mention, is that if that "V" feature is the most important, make absolutely sure the spinal bend curve is the right length (equal to the flat length otherwise it'll stretch or compress the solid - I use relations and a perimeter dimension to ensure that), and it'd be best if the curve used is at the point where the "V" surface and outer surface of the cylinder intersect. A spinal bend will compress (shorten) material on the inside of the curve (essentially the neutral axis) and lengthen in on the outside. Whereas actually bending sheetmetal will pretty much keep the same length on both inner and outer lengths, making the end surface not normal to the spine arc, like it will be with a spinal bend.

Re: Inaccuracy with Spine Bend

Thank you Frank,

I had the suspicion that I should be sure the distances matched but wasn't for sure whether this was necessary.

This is good to know.

Re: Inaccuracy with Spine Bend

Hi Frank,

I just wanted to let you know that we enhanced Spinal Bend in Creo 3.0 so that you have the option to keep the length of the original solid now, instead of having the system adjust it to the length of the spine.


Re: Inaccuracy with Spine Bend

Sounds like there's some good enhancements to it. If, for instance, you only wanted to affect the middle of a part, you'd have had to model as solid the center, then do a 'bend on it, then model the ends. Or, as I've had to do, make it a quilt, bend that, then solidify that.

What "original length" does it keep, wherever the neutral axis (where the spine curve) is? Like I mentioned, actually bending a solid, the length should stay the same on both sides of an object, thus the inner cylinder would rotate a few more degrees than the outer, making an end surf that is ot normal to the spine.

It's one of my favorite tools, so it'll be interesting to try it out.

Re: Inaccuracy with Spine Bend

Bending about the neutral axis should not keep the inner and outer lengths the same for almost all materials. The only case where they remain the same is when there is no resistance to shear, such as is seen in bending a deck of playing cards. Without shear there is no stress and without stress, no force is required to make the bend. Does it require force to bend playing cards? Sure, because the individual cards have a thickness and develop shear, though it is very small in comparison to the case where all the cards are glued to carry shear loads.

Re: Inaccuracy with Spine Bend

Shouldn't in theory....but generally does. Try bending a piece of tubing 90deg or 190deg, you'll certainly see that while material outside of the neutral axis stretches somewhat, the material to the inside of it does not compress. Way back in the day 9early '80's) when I was pursuing a welding career, we'd have to bend 3/8" or 1/2" plate 180deg to test the welds, and there was a definate angle to the ends. It depends a lot on the material, and how it's clamped when bending. If it's clamped tightly, the ends are perp to the neutral axis, but the bend thins out. Also, say, in sheetmetal, the actual bend is tight, and very localized. This looks like they roll a plate.

Re: Inaccuracy with Spine Bend

I guess you are answering something different than what I wrote. Generally things work according to theory because otherwise no one would use the theory.

Tubing, unless it is supported internally will collapse if bent far; tubing is often stretch formed over a mandrel forcing a displacement of the bending neutral surface away from the neutral axis, even outside the part. In this case both the inner and outer fibers are longer than they were originally.

Weld test samples are very short compared to the length of the bend and are very thick relative to the length which will distort the end faces. Even so, the inside length post-bending will be less than it was originally, unless it was also stretched. Check slide 15 of PPT – Sheet Metal Forming PowerPoint presentation | free to download‌ to see the difference between bending and stretch forming. Stretch forming would require negative K-factor to model more correctly.

If you look closely at bent weld samples you will see the sides flare out on the inside of the bend making the part wider because of the compressive load on the inside face. This will happen even if the bend is started with bending press or a 3-point bender and finished by finishing by forming in a vice. Yes, the neutral surface shifts as the outer fibers thin under tension and the inner fibers thicken from compression.

Unless it is stretch formed the inside surface gets shorter and the outside surface gets longer.

What is more applicable to this case is that one can't roll-form to the end of the part. To do so requires applying a bending moment with decreasing leverage, which results in a requirement for an infinite amount of load; not realistic. It needs to have a straight section lead in and lead out to represent the amount that does not bridge all three rollers or the three faces of the bending dies.