Too many iterations for the feature calculation

[[Ri...]]

I am getting this error on a large diameter cone (15+ inches in diameter), its similar to a V groove pulley.
"Too many iterations for the feature calculation"
What does this mean?

Of course only one cone, which is simply the opposite side works fine.

[[Da...]]

I usually get that when there is some major form error. Could it be shanking? Try reducing the number of points.

[[No...]]

Technically, this error means that Calypso was unable to calculate a stable result within the maximum allowed number of iterations. There must be a maximum because sometimes the best fit algorithm can't find a result and would be trapped in an endless loop. Elements most susceptible to this problem are cylinders and cones. Possible reasons are: suboptimal distribution of points on the cone, a bad diameter/angle/length ratio, the cone not having a full 360 degrees, etc.

[[De...]]

I will preface this by stating outright this is my basic understanding of whats happening and probably is not totally correct as I have no inside knowledge of what Calypso does internally, however based on how the math of how fitting works I think its not totally incorrect.

This is caused when the data is not making sense to the algorithm.

The algorithm attempts to determine the location and size of the feature by comparing the actual values to the nominal values defining the feature.

It uses a iterative analysis of the points based on the nominal size and position of the feature and standard deviation of the actual values to determine the actual size and location of the feature. If the actual values vary enough from the expected nominal values, the comparison does not yield results that are consistent and this error crops up.

This can be caused by a number of things, the following are some common examples.
Bad form of the feature being measured, measuring too small of an arc segment, measuring a cylinder or cone whose diameter greatly exceeds its length.

In all of these situations, the data collected is often not clear enough to provide the necessary information for the algorithm to properly determine the location / size of the feature.

In order to get better results, the best bet is to make every attempt to collect the best data possible, on a diameter for instance make sure your scan paths are at either end of the diameter as much as possible and possibly add additional paths.

Sometimes making an alignment to the feature prior to measuring can make all the difference, however sometimes the feature to be measured is just asking too much of the CMM.

[[Ri...]]

Its a simple "V" on a large diameter.
One Cone calculates fine (Left side) the other (Right side), with exact parameters and Stylus, fails?
This doesn't make any sense.

[[Ow...]]

Turn the filters/outliers off and see if it calculates. If it does then, "there's your sign"...lol
Seriously though, it may help understand what's being calculated incorrectly.

[[No...]]

Please sign in to view this quote.

It's quite correct, I'd say. I intended to write something like that too, but then I decided to make it short. Not that I know all the math behind best fit. As far as I understand it, the algorithm looks for a remaining error that is below a certain threshold. Normally the error gets smaller with every iteration, but sometimes it does not. I may be wrong here, but I think I read somewehere that it can even oscillate between two or more unstable states and never finds an end.
So when the remaining error stays above the threshold or the algorithm is trapped in an oscillation, it stops at the maximum number of iterations and throws this error.

That one identical cone is working and the other is not doesn't prove much. Sometimes it's a very small deviation that tips the scale. Calypso has improved a lot here, but in UMESS we had a number of parts with cylinders in serial measurement which worked 99% of the time, but every now and then one part failed.

[[Ri...]]

Thanks for the replies!

[[je...]]

I've found that adding a evaluation constraint to the normal vector helps in some cases.

[[Re...]]

When possible I will do a helical scan on my cones. This scanning option on cones has greatly reduced the number of times I get the dreaded "Too many iterations". From there I would reduce or increase cone depth. If that doesn't work then I start reducing the number of points. If I have a cone that cannot be measured with a helical and there are circle segments, I will make sure to group the circles at each level. Right or wrong these are the little things that seem to get me though the too many iteration issue.