hight point plane

[JEFMAN]

I constructed a high point plane from 32 hits. 8 hits are higher than the plane. I calculated a tangent plane with Excel that give better results...
For the two cases, the barycenter of LS plane is included between three points selected by the calculated plane.
Any explanation on high point method would be great !!!!

[AndersI]

Reading the PC-DMIS Help File on High Point Plane makes me think someone didn't know what they were doing...

I have so far used a process of my own, first calculating the BF plane, level to that, find the highest point and construct a plane parallel to the BF plane, through that point. But this is more of a tangent plane than a 'high point plane'.

[DJAMS]

I have so far used a process of my own, first calculating the BF plane, level to that, find the highest point and construct a plane parallel to the BF plane, through that point. But this is more of a tangent plane than a 'high point plane'.

This approach has a lot of merit - It is a good compromise between "right" and stable. My experience with true "high point" algorithms has not been good, using numerous softwares over the years.

[EMacDonald]

I have been unable to get HP planes to function at all. I start with a collection of measured vector points, but the HP plane constructs are nowhere near the points. Sometimes an error message for "cannot construct using present points" or similar shows up, but usually the theoretical and constructed planes are several thousand (!) mm from the nominal location. BF planes work as they should, with rational constructs, based on inputs.

[Rondog]

I constructed a high point plane from 32 hits. 8 hits are higher than the plane. I calculated a tangent plane with Excel that give better results...
For the two cases, the barycenter of LS plane is included between three points selected by the calculated plane.
Any explanation on high point method would be great !!!!

This isn't really a reply but more of a question.

With a high point plane there is only one answer - the plane using the three highest points. If you look at all of the points relative to the constructed plane they should be on the material side. In PC-DMIS, if you level and zero the plane then ask for the position of all the points they should all be on the low side.

How did you check this in Excel?

[AndersI]

With a high point plane there is only one answer - the plane using the three highest points.

Highest relative to what? Current workplane? BF plane of all measuring measuring points? Something else? What does the standard say?

And what about pathological cases (think a round flat plane with a 'spike' in the middle and all other points good - there is an infinity of 'high point plane' solutions for that configuration, with very different directions).

[JEFMAN]

In PC-DMIS, if you level and zero the plane then ask for the position of all the points they should all be on the low side.

You don't need to level on the plane to see that it doesn't work...
ASSIGN/V1=MAX(DOT(PL1.HIT[1..PL1.NUMHITS].XYZ-PL1.XYZ,PL1.IJK)) with PL1 = high point plane gives (in my example) a value of 9 µm for a flatness of 30 µm. If you don't assign max, but only dot product, it gives an array that shows all the values above the plane (8 in my case).

[JEFMAN]

Highest relative to what? Current workplane? BF plane of all measuring measuring points? Something else? What does the standard say?

And what about pathological cases (think a round flat plane with a 'spike' in the middle and all other points good - there is an infinity of 'high point plane' solutions for that configuration, with very different directions).

You're right, there are many solutions, but ISO std says that you have to choose the tangent plane, free side for part, that minimize the max distance (Tchebychev), so you have to calculate this. It comes from classical metrology (marble ?), but it would be better to take LS plane in standards...

[JEFMAN]

How did you check this in Excel?

You can do it with "solver". You just have to calculate the distance between each point to a plane (for example LS plane), then minimize the max distance by constraining the 3 min distance at 0 (so the plane is tangent). It's not very difficult, so it's hard to understand why PC-DMIS doesn't do it !!!!!!!