Nailing out Filter/Outlier Settings.

[[Mi...]]

I'm trying to find the balance of filtering out dirt, without overdoing it. I think because we use the highest point density available for the given scanning speed, we are ending up with more data than the cookbook settings were envisioning.

EG. this is a 25mm diameter hole, scanned at 10mm/s with a 0.04mm step width totaling around 2000 points. 10mm/s is the limiting factor so the thought was that more data never hurt so take the most points you can at that speed. But I think it actually is causing some problems.

Here is an example of a hole scanned, first with no outlier, second with Cookbook filter/outlier settings. They had to recheck this part due to the size coming up small. It clearly scanned over some dirt, but the standard settings don't quite filter everything out.

I can either add an iteration, change the factor for outlier (maybe 3 inside, 2 outside), or increase the number of points removed with the outlier. All of these options fully removed the erroneous data, I just worry than I'll end up filtering out too much data and miss a potentially out of spec part. These parts are too important for us right now to potentially miss something like that.


Thoughts?

[[Ma...]]

I don't think filters will help much, but did you tried remeasuring it? If you can not clean that before run, then loop could help with dirt - but no for solid chip.

[[Mi...]]

Please sign in to view this quote.

Yup, after a reclean and remeasure the hole is fine.

Thats why I'm hoping adding an iteration is acceptable. Here is the same data, cookbook settings except the number of iterations changed to 2. This looks fine to me, and the diameter now matches the recheck they actually did after recleaning the part. It also affected the results of the BF position as this hole is part of a pattern.

I just don't want to stray into the "Filter/Outlier it until it's good" category

[[Me...]]

Why not just clean them before measuring?

I tried to clean every part before running.
Nothing worse than running a 20min program and a spec of dirt puts it OOT. haha

[[DW...]]

Please sign in to view this quote.

Nothing worse than getting dirt on a stylus system, not catching it, and then throwing off all the measurements down the line.

[[Mi...]]

Please sign in to view this quote.

Is that a serious question?

I don't run these parts after I prove out the programs. There are 33 Zeiss CMMs in our facility ran by dozens of other people, myself and one other guy program.

Yes. We tell them to clean the parts. Often enough they don't get it clean enough, then time is wasted on rechecks and sorting.

Managers then tell us to make the CMM programs better to prevent rechecks because the time isn't invested in training. Some parts we use PCM to automatically rescan holes based on the form where they've been problematic.

None of that is actually relevant to my question though about outlier settings beyond standard cookbook ones.

[[Da...]]

Your parts look really round, are they ground? Outlier elimination is great. We set for 5 adjoining points and 2 or 3 iterations. Each time an iteration is run, the next iteration is run with the remaining points, so it actually decreases with each iteration. We used 1500 to 2500 points. Manually checked with air probing afterwards to verify. It works quite well.
Our guys would use those godawful red shop rags to "clean " their parts! Lint was our biggest problem.

[[Ri...]]

I've been on both sides of the fence.

Outlier Elimination is the biggest gray area in measurement. How do you know the data you removed was dirt/debris? How do you ensure it wasn't a gouge or flash material?

Outlier Elimination on when the Evaluation is LSQ shouldn't make a huge impact unless your point density is rather low. Outlier Elimination on a Form characteristic or a Functional check can potentially make a bad part good.

Scary stuff.

I understand the need for it, especially in a production environment though.

[[Je...]]

Please sign in to view this quote.

.
Hi Michael. Great questions on filters & outliers.

After reading all the info you provided, I think the CMM and software is doing what it should. It really does seem to be an issue of workpiece cleanliness, and if the operators don't want to remeasure parts, they may have to clean a little better. My apology if this isn't a well-received idea.

You could also LSQ and filter the snot out of the features, but with tight-toleranced critical features, this would be a risky strategy.
Stable, clean workpieces are essential.
.

[[Mi...]]

Please sign in to view this quote.

My exact worry. This diameter isn't LSQ, its Max Inscribed. This pattern of holes has a 0.1mm BF positional tolerance with only 1 degree of freedom unconstrained - and it's a customer that actually understands the ASME standard who will verify the size, position and perpendicularity of these holes all at Max. Inscribed for the functional fit. Meaning we need to do the same in the production environment.

[[Mi...]]

Please sign in to view this quote.

I wish they would burn all the red rags...

The roundness and surface finish is impressive on these holes, but they are just drilled and reamed. an average part has about 0.003mm of roundness in a given hole.

I'm really tempted to throw a second iteration on... but the customer just returned parts that had a spiraled score mark in the hole that the CMM already filters out with standard settings.

[[Mi...]]

Please sign in to view this quote.

You're not wrong, cleanliness is absolutely the root cause. We have big warnings about it at all the CMMs, some programs even stop and yell at operators if things aren't clean enough especially on datums.

I'm just trying to see if there is something more I can be doing on the programming side to help reduce the impact of laziness in the production environment.

[[DW...]]

Please sign in to view this quote.

This is at best a misunderstanding of how CMM programming works, and at worst purposeful negligence on the part of your managers.

Short of using PCM to execute a popup asking the CMM operator if they have cleaned the part, there is nothing you should be doing to massage the results because someone couldn't be bothered to clean the part correctly.

Down the road when these parts are being scrutinized after a customer rejection, the over filtering will come into the light and your managers will no doubt conveniently "forget" they asked you to make the programs "better". Again, either because they have no idea how the software works or because they are willfully negligent.

Working with production usually requires some level of adaptability and compromise. There are programmers who will not budge on their strategies, and others who will bend to every manager and set-up guy's request. I suggest you fall somewhere in between. This is not one of those times where I would bend. Best of luck.

[[Me...]]

Please sign in to view this quote.

Hey man,

Sorry If that question if that was rude, wasn't intended to... was just curious.
In my experience accuracy is achieved simply cleanliness.

Lol Your "Management" sounds just like a place i used to work. 🙄
Definitely a gray area. Just put Programable stops every feature that say Did you clean this? 🤣

[[Da...]]

Does your crew have supplies to clean properly? We had alcohol and several different sized bottle brushes. oh, and an illegal air gun extension tip. Procedure was trained and retrained. (I only had 4 machines and 2 operators)

Filtering really doesn't do much to outliers. Keep your filters set to the CB. Outlier elimination can do a lot. Use fewer adjacent points (2-3) and up the iterations (2-3)

I dealt with honed or ground holes so outliers were almost always dirt. The process can't produce ID "bumps" very well. The process produced holes with 1 micron roundness. LSC was acceptable and made for very repeatable results.

Play around with it. I like your idea about from 3 inside and outside to 3 and 2 outside.

[[Mi...]]

Please sign in to view this quote.

Yup, Alcohol, cotton swabs, Lint free wipes, brushes, air guns with extra clean air, instructions on which order to use everything on the desks.

Our process will practically never create "raised" material on a hole like this, but it can create scoring into the material - that's where my idea of having an uneven factor might help catch at least some more of the issues that result from dirt, without unintentionally removing too much data and risk not capturing an oversized hole.

We also have to measure the Min. Circumscribed of these bores to give a "MAX" Diameter, because the customer presses a pin in, and they will send complaints about low press force and the Min. Circumscribed result pretty closely correlates to when they will start having issues.

[[Mi...]]

Depending on the depth of the hole, is it possible to create a second scan path, and apply filters that way?

The thinking is with the second scan path hopefully far enough away from the affected dirty area would be able to give you a good indication of its actual form, that way you can know if the feature actually does have a gouge/burr/defect in that area. If you have a large difference in form, then the operator should check if its dirt or an actual machining defect.

[[Er...]]

Maybe...

have examples of "good, bad, ugly" and make adjustments so all measure accurately.

add scans with lines to catch scoring.

have a fixture with air nozzles to clean the critical features of the part.