Best fit or iterative alignment?

[Jay_Q]

I have a part around 3ft long by 1.5ft wide. No flat surfaces all curved. The customer originally had 6 target points which I did a iterative alignment to. The problem is the part is twisted and out of position so bad that it wasn’t taking the greatest hits along a thin flange around the perimeter. Results were questionable. Then the customer says don’t use the targets just best fit the whole casting and report the profile. Still the problem was getting good hits. So I started with the 6 point iterative then in dcc took 9 points in z, 3 in x+ 3 in x-, 2 in y+, 2in y-, then I 3D best fit these 19 points and took 800 more points then best fit the 800.It seemed to take better hits but the profile results from the 800 points was way worse than the 800 point from just the iterative. Long story short with a large twisted part what’s the best “pre alignment “ to take good hits to do a best fit profile?

[AndersI]

If the part is twisted enough, no simple alignment strategy will be enough to collect all the hits in the correct places. Relative measurement, edge points, long prehit distances etc. may help, depending on the form (and form error). Different alignments for collecting points on different areas of the part might also be a way.

[Jay_Q]

I have 2010 software, I don’t have edge points or relative measure option. Multiple alignments seems like it could work. Does it matter what alignment I create the scans in? Do I collect all the data using multiple alignments then best fit all the points and then create the scans? Or does it not matter? I’m just worried about it not fitting it properly. Because of the different nominal from different alignments. Most the parts I check have datums I don’t best fit much.

[AndersI]

Edge points and relative measure can [with a little effort] be emulated by local alignments. Alignments used for *driving* to the correct place in the correct direction can be done in whatever way you find is best, but the alignment for *dimensioning* must be according to the print - if it is a "surface profile, form only" you do a bestfit on all surface points (that have been measured in the correct way, for example ignore manual points made for finding surfaces and such).

[Matthew D. Hoedeman]

VECTOR points should be able to be done "relative to" another feature (not exactly sure when it was added, but it was at some point, before that, surface points could be relatively measured). Not sure that ANY best-fit or iterative alignment will get you there with a twisted, warped part. If, for example, you are having trouble getting a point on a short flange because the part is warped 'down' so the probe passes over the top, simply take a hit on the warped surface, then measure relative to that point for the flange. If you truly don't have the relative option, then simply do a 'quickie' alignment, origin Z on the surface, measure the flange, then go back to the real alignment. This would suck for a lot of points, thus the RELATIVE TO function.

[brian.freeman]

I have used this type of strategy before. Make sure you insert command RMEAS/LEGACY (I,J,K, X,Y,Z), you will get better results.

[Douglas]

I have 2013mr1 and it will do 'relative to' in auto features. For measured features you would need to assign the edge point hit to a variable and use that.

I do grooves this way to probe near a wall... point on the wall assigned to variable then the theo z might be V1+.025

[Jay_Q]

Thanks everyone. I use relative all the time (using alignments). As Matt said it would suck to do this with the amount of points I need to take. I was hoping I was missing something easy on the “driving” alignment. I was also worried about multiple alignments “confusing” the best fit as I don’t use them for surfaces much. I am clear now on what to do. Thanks again, you guys are great!