No smooth edge on loft of propller blade.

Hello, I have been struggeling with a propeller blade before, but that got fixed using the forum. That blade was kind of tapered, now I'm working with a blade of constant chord length.

When I create a loft over all my cross sections, I do get an ugly edge, I cant find how to make this looking better.

The first part looks nice, but the further I get from the start the worse the edge. If I do start from the other side the loft looks even worse, probably because off the larger angle between the cross sections near the hub.

(Using Autocad 2011)

What can I do to improve the loft?

With kind regards,

Edwin

I attached the cross-sections as splines.dwg for the ones who want to try.

splines.dwg

I'm not a loft professional, but the trailing edge probably shouldn't "go to zero" like that, I suspect Loft needs a thickness there, since it's trying to work out how to expand and compress the shape to reach the next crossection.

If you put a rounded edge there you might have more luck with it. I'd at minimum consider 1/32 of an inch radius (.03125) instead of that .008 dimension in there now.

Ideally you would have only two guide curves (you have none) and only two profiles (a beginning end ending). Fewer profiles is better. And your profile splines have too many nodes. I don't have the tools in AutoCAD to demonstrate why, but fewer nodes in a spline is always better. The best spline is only two points (and yes, in other software a smooth curve - not a line - can be generated from two points). I'm pretty sure I wouldn't fool with AutoCAD for prop design.

When I create a spline I start with 2 nodes and adjust curvature (not in AutoCAD) and then add nodes only as needed to get desired curve.

JDM: Would you mind expanding on your comment, for my education, regarding "fewer nodes in a spline is always better"? Thanks.

Hello everyone,

thank you for the reactions sofar.

The cross sections in the design are created using an other program. I have traced these with a spline, ending in a curve on the trailing edge.

Every cross section is slightly different, and angles between the cross sections get smaller towards the tips of the propeller.

When I use twoo ends and a guide I get a strange model. When I use all cross sections and a guide. (guide at the end of my splines, near the trailing edge) I get a nice trailing edge. But the leading (round) edge now looks weird near the hub (thickest art and biggest bend).

I guess I should try to create an extra guide at the leading edge.

Edwin

JDM: Would you mind expanding on your comment, for my education, regarding "fewer nodes in a spline is always better"? Thanks.

I will try to explain best I can (easier to show with actual parts, but I don't have anything at hand right now).

The software has some algorithm (I don't pretend to understand the math) that calculates a best fit curve between points (actually there are several different fit methods you can choose in AutoCAD).

Imagine that you knew the formula and were manually calculating points along the desired curve. How many points do you calculate? How many decimal places do you carry out each calculation to place the points.

For manual calculations the tendency it to assume more points is better - and probably it is.

But how many points is the computer calculating internally for the same curve? My guess is several thousand more than you care to calculate and probably out to more than 8 decimal places.

So in CAD specify as few points as possible and let the software interpolate between the points.

The attached animation shows spline curves with something called a curvature comb. The curvature comb is a graphical method of viewing the "smoothness" and change of curvature in a curve. Just by visual impression the curve from two points seems to be smoother. In fact it is very difficult to get a smooth curvature with more points.

Here is an example (I wish I could find the actual file).

A customer had a coordax (coordinate axis measuring machine) generated set of points to follow a ship hull. They made mating parts to follow the curvature of the hull. They tried to argue with me that they had to use the generated points from their customer for accuracy.

My first question to them - what is the accuracy of your manufacturing process to the given points. Turns out their process is not all that accurate (we are not talking precision machined parts here).

So my second question was, so we can't toss about two thousand of these points (literally) and instead find a best fit curve that is still well within your manufacturing tolerances?

BTW in that video those grips on the ends of the splines are called Handles (straight line with a grip at each end of handle) - there is also another curvature control that I haven't shown here. The handles control "weight" of curvature.

Artsy-fartsy (technical term for ID people) industrial designers get real concerned how the geometry reflects light. This 2D example has similar tools for 3D analysis of the resulting geometry.

Thanks JD.. very informative.

Again.. I'm not a "lofter" but I have a question, JD..?

Would you say that each profile used in a multi-profiled loft is taking the overall transition and then spliting it into transition pieces, or would it be a step-by-step transition down the length.

Ie- would they get a smoother result merely running loft from profile 1 to 2, and then 2 to 3, (rather than 1 through 12) or would it start looking segmented?

The reason it might look segmented would be because the loft would have no regard for what the curve was doing before it got to the starting point.. It would start off and end a fresh curve every time.

I suggest that because of the "ripple" effect I see in the edges of the OP's image.. it looked like Loft was compensating for a pre-entered curve (the previous curve) which created a "lump" when it hit the next profile, and it again tried to compensate, but seemed to over-compensate that time, creating a dip.

**Additional**

That "ripple" is why I was suggesting they should thicken the edge..

Perhaps instead, they could scale the whole thing up. This would make it an easier set of numbers for AutoCAD to generate, and avoid the effects of "rounding numbers up or down" when they hit decimal tolerances of the 20th decimal place or whatever. That might alleviate the ripple.

Then scale the whole thing back down again when the loft is complete.

I fooled with it a bit in AutoCAD and decided it was a lost cause. I don't think section by section is the way to go.

I thought about fooling with it in Inventor - but the curves don't import as cleany as I would like and I wasn't motivated to recreate them.

Assuming the original points came from an *.xls file I might be motivated to experiment a bit more given the xls file....

See the red "seam" in this solid. It actually curves around to the other side of the part at the left side in image.

My guess is that this seam connects the first point selected when creating each spline. (hmmm didn't one of the MDT tools make it down into vanilla AutoCAD 2010 to check the start/endpoint and direction of a spline curve - I have some checking to do)

A couple of tips I would do - make sure each spline starts from the same relative postion along the solid. Make sure each profile spline has exaclty the same number of points (lofting does something called point mapping where it connects a point in one profile to a point in the next porfile - in Autodesk Inventor you can manually control this point mapping if the automatic solution isn't correct).

ohh I almost forgot, set facetres = 10

I do now have got Inventor installed, but I guess I need a lot of tutorials to get around in inventor.

All the splines are traced from a dxf file, like the one you have seen before JD, when I could not get a fillet on the curved edges.

All splines have the same number of nodes, and start at the same postion.

When I loft only a few splines, the images looks even worse. (see below)

When I loft only twoo splines, it looks like the spline starts perpendicular to the cross section, maybe this is why the edges look like this.

If I would have a guide on each edge of the profile, i might get a nice loft, unfortunately I don't know how to get a good curve at both the leading and trailing edge.

Edwin

Interesting.. this is actually a problem whose results seem to match the twist issue another poster had earlier. The "sweep" where the starting angle of the two wire profiles needed to be tilted so that they "grew" as a twist, rather than slide sideways as they rotated around the guide.

Unless I miss my guess, loft is "sliding" the pieces together, rather than rotating them. The result in that last image shows what would happen if you extruded a part and rotated it 'too fast" as it grew, rather than tilting the profile at a slight angle.

http://www.cadtutor.net/forum/showthread.php?t=47210&highlight=sliding+turn

The pieces need to rotate as they loft, which is why that red line in JD's image goes from the edge to the back of the part.

This thing needs to sweep twist, but also grow as it goes up..

Like.. loft from beginning shape to end shape without the twist, and THEN add the twist to it somehow... Is that possible?

What happens when you loft it, using only the start and end shape? (with the twist, and without)

**Update**

I think I've got it.

To accomodate for the fact that loft IS twisting/rotating the profile, and therefore screwing up.., take your drawing and look at it from the "Front" 3d view. From that aspect, make a flat grid. Make the grid whatever you want it to be to reach the tolerances you need. Then use that same grid for reconstructing each blade profile (intersections between shape and gridlines). Don't rotate anything..(especially not the grid) just reconstruct the shapes so that the start points (and all subsequent points) are at one position in relation to that grid. (ie- the leftmost point of the propellor profile perhaps).

This will stop loft from failing to twisting/rotating the shape correctly and concentrate on simply reshaping it as seen from that single angle position, as it's intended to do.

If you need a single point to be consistent all the way through (as I think you do), you may need to do the blade in two halves so that the central rotation point BECOMES that single point. Then loft and then union the two halves together.

Hello Mike,

I've been sruggling to understand what you mean.

In the meantimeI constructed some lines to help me get some inersections on the leading edge.

On these intersections, and de endpoints of the cross sectios I drew twoo splines to have two contol paths which control the twist in de blade.

I created a loft using five cross sections, both ends, and tree in between. This gives a great loft.

The blades thicken more nearing the hub, this is why I used not just the ends, but chose a few in between.

With kind regards,

Edwin

I get that a lot.

I'm glad you got it working.

The basics to what I was saying was.. Loft isn't doing a good job of rotating, so don't try and get it to do that. Stretch/warp a copy of the original profile into the shape of the next profile, rather than rotating the original profile and adjusting it to make the next profile.

That's why the red line in JD's image went from one side of the blade to the other.. if you look at the prop from the Front view, it was simply lining up the profiles, so I was suggesting you cater to that, since the formation of the profile is controlling how loft deals with it.

Anyways.. that probably didn't make sense either, but I'm glad you got it working so I can stop thinking about it now.