airfoil curve scan issue

[[Ba...]]

I scan the airfoil of a integrally blade rotor.
at the leading edge, there is a high spot, I have measure it several times, it still a high spot.
and I displayed the measured points, I think the measured points is the mid point of the stylus sphere, is it right?
I think if the blade profile has a high spot, it should also a high spot on the measured points. but I can not see high spot on the measured points. how do you think?
the leading edge radius is about R.008 inch, can CMM accurately measure it?

[[An...]]

is the high spot the start or end of the scan?

[[Er...]]

A cmm with an active sensor can absolutely measure this accurately. And a passive one too. Even if the curve deviates from nominal, just takes some more elbow grease.

I can see a very common issue with your nominal data by looking at your point distribution.
When you go from Pressure/Suction side to the high curvature area at leading trailing edge. You have a "big gap" in nominal points. This WILL make it harder for the software to calculate the curves mean line. And that will carry over errors on pretty much anything you evaluate from your curve. Form is not affected by mean line tough. Try to distribute points more like this picture.
In my opinion you don't have a sufficient number of nominal points on the leading edge either. You will need more to accurately define the nominal curve.

Another thing that comes to my mind is, where do your segments start and end? At the tip of Leading-/Trailing edge? How much do you overlap? You should always use filtering when scanning, noise will occur.

R0.008 isn't that sharp, so it should not happen, but. Some times when I measure blades, the leading edge is razor sharp. That can cause the probe to "stick" right at the high point. There are three things that usually fix this:
1.) Sort the points, (Wont fix the sticking issue, but can fix your data.)
2.) Measure the curve, Create a WP-System of the curve (Translate/Rotate is axis suitable for your application), Re-measure the a copy of the first curve placed in the WP-system. (This will compensate for where the blade geometry actually is located in space. So the CMM follows the surface better.)
3.) Singe points. Some times when sharp blades have a lot of "twist" and "lean". Nothing else can fix it.

Measuring blades is a science in it self. But try the above mentioned. If that don't fix it. We can try some more advanced approaches.
Out of curiosity, what are you evaluating?

[[An...]]

if you use a Gaussian filter, the end point must run beyond the start point, as the Gaussian filter requires a lead and trail

[[An...]]

Please sign in to view this quote.

Using spline and intersection point should solve the "Curvature Problem".
Watch the magnification!
See attached.

Highspot_and_Airblade_1.pdf

[[Er...]]

Here's the approximation error of the transition area from line to arc of a slot with 100 points. Gives a quite clear picture.

[[To...]]

Please sign in to view this quote.

Wouldn't Spline filter be appropriate here?

[[An...]]

Please sign in to view this quote.

A spline filter is definitely the best choice.
I do not know which filter "Baishuo Lio" used!

[[Er...]]

Are you shure Andreas? If you have steep changes in the curvature, Gauss is the way to go. And from Pressure/Suction side of a blade to leading-/trailing edge, there is a usually a very steep change in curvature.

[[An...]]

Leading and trailing edges are not my science.
(Not an employee of the LUFTHANSA)

I would say that the use of splines minimizes in a certain manner
the lack of nominal points in the area of curved edges.

Curve_and_Spline_1.pdf

[[Er...]]

Oh, now I understand what I didn't understand. (Some times I have bad luck when I try to think.) Yes, indeed it can decrease the deviations by approximating. How ever, leading edge is most crucial for the expected airflow. It can cause unwanted turbulence along the rest of the profile with loss of effect as result.

So, if you only wish to get good results. Go with spline.
If that is going to fly, and I might end up on that particular flying object. Try some of my first suggestions. 🤣

[[Ba...]]

The high spot is not the start and end of the segment.

yes, Eric, I have read the instruction of Blade Pro, it said the nominal point density must be chosen according to the curvature, and should also increase the point density at the pressure and suction side near the edge, just like the picture you post.

I guess the nominal point I used has too much density at the pressure and suction side, it is useless. But the density at the edge is not sufficient, the scanned edge profile looks not smooth. Is it?

And, what is your criterion for generating the nominal points, use chord height .00001? or step width xxxx. How long is ok for picture below?
Below picture is the segments start and end point, overlap only on this one point. if there is overlap, then 1 nominal point will have 2 actual point, how calypso choose?
I did not use the filter, I am not familiar with filter. Calypso instruction said the low-pass filter can eliminate the effects of surface roughness, but I just guess the roughness is very small, it will not affect the evaluation of the profile tolerance. What’s your understanding about the filter, and how I set the wavelength? Do you have some practices of filters?
I still confuse that why there is high spot, it is higher than adjacent points very much, but I do not think the actual blade is like this.

Sort the points? Which one is it of the picture below, Sort measured points changes the point sequence. Approximation can smooth the measured points or nominal points. I have not used these 2 options.
What I evaluate the curve is the TP of innermost section stacking point, TP of outermost section stacking point in relation to the innermost section stacking point, TP of intermediate section stacking point in relation to the innermost and outermost section, max length, chord line angle, thickness at a basic distance to LE and TE, airfoil profile, LE TE profile or radius, smallest radius of LE and TE (in blade pro instruction, it computed smallest osculation circle), flat and reversal on LE and TE. And I have some question when evaluate these items.

Eric, you mentioned decrease the deviations by approximating, is it the Approximation in the picture above?

[[Ba...]]

Andreas, how you generate this line and what the 1.5mm mean.
I don't use mm, I use inch.
"I would say that the use of splines minimizes in a certain manner the lack of nominal points in the area of curved edges."
How to use spline, do you mean the caplyso will link all the points with spline instead of straight line, I am not understand your idea.

[[Er...]]

So many questions, so little spare time to answer... 🙂

Well, I wouldn't call the density to high of the pressure/suction side. It seems to be good. But you see the "gap" of points that occurs between the spline entities of the curve at the start and end of the leading edge. It's larger then your distance between the points that you allow. You need to set the gap to 0 when creating points.

Where I work, we define the size the leading edge like this:
When a circle of perfect form tangents the curve and the distance to the LE nose index (perpendicular to the line created by the tangent points) is equal to the diameter of the circle. It's not easy to calculate in Calypso tough.

How ever, depending on blade geometry this can be anything from 0,005-20% of the chord.
Here is a picture of a random blade profile. As you can see, the density of points continue past the extreme curvature.
This works great for us. But we need very many points since we have created more evaluations then available as standard in calypso with pcm and python.
You should always filter when scanning. And there are no easy answers on settings. We have many hours of testing in our filters. They are based on expected surface roughness, scanning speed and probe diameter. But standard filers are better then nothing I guess.

I was referring to "sort measured points", That can solve issues when the probe sort of gets stuck on sharp edges. Making for example point 45 occur before point 44.

If you filter by spline, an "arc" will be calculated between the points, making it smooth. But the surface might not look like that arc. So Calypso is guessing the geometry. More points is the way to go if you want accurate results.

[[An...]]

Indeed,the point density influences the result.
Regarding "NURBS"-curve leads to amazing insights.

[[Ma...]]

I've always changed the amount of measured points to 3x the amount that I use to define the nominal. I.e., if I section the model and use 100 points to define the curve I open the strategy and change the amount of points to 300. I think this give Calypso a better advantage of triangulating the nominal points.

I would also agree you may not have enough definition around the trailing edge. I would redefine the curve making sure I choose a minimum of 100 per section then then change the amount of points to 3x the given value.

Mark

[[An...]]

Parabola and Spline.
See attached.

Parabola_and_Spline_1.xlsm

[[Ba...]]

Eric, I seen this at the calypso cookbook:
If filtering is active for the feature, make sure that all scanning paths on
the feature have the same step width or the same step distance. Otherwise,
the results are not in accordance with standards.
It is not easy in airfoil scanning, because the measured points step width at the LE/TE are much smaller than other area, so, is it a question? what is the "the results are not in accordance with standards" mean?

And I seen this at the calypso Kurve cookbook:
As the spline filter is only used for the measuring points and not for the
deviations of the measured points, larger deviations may result from the
distance evaluation. This especially refers to the spline ends with open
contours and areas that are subject to a high degree of curvature
changes. Do not use the spline filter in such cases.

is the filter is only used for the measured points(before the probe radius correction)?
why it said do not use the spline filter in areas that are subject to a high degree of curvature
changes? I don't understand.