Filter/Outlier

[[wi...]]

I have a Ø40.014 ID (+.011/-0) That I am scanning.

We never scan and have always used a 5 pt method

I have my angle or measure 450°.

Do I need to adjust the Speed for this characteristic? If so what/where?

And as far as evaluation and outlier I was looking through the cookbook and have it setup but curious what you would use as far as these features?

I have differing roundness between the scanning and 5 pt methods which is expected. However some of my traces are showing perfect roundness with scanning and some are showing deviation. Just looking for some guidance as I pick this method apart.We never have scanned since I've been here and our sister companies always use 5pt.

Thanks for all who help I am doing testing since we have time and im curious.

[[Ri...]]

How you would ever get by with 5 points is beyond me. I guess it would detect common shapes like egg or tri-lobed, but wouldn't paint the entire picture.

Let's just assume this is a short Cylinder, so let's measure it as a Circle:

Per the Zeiss cookbook the settings are as follows

- Filter should be 150 UPR
- Outlier should be Sigma ±3
- Factor for Outlier should be 5 adjacent points
- Pre-filter should be 10-5000 UPR

# of points should be at minimum 1,270 for 380°. The formula for calculating number of points is based on the UPR Filter you are using multiplied by 7, so 150UPR x 7 = 1,050. That is based on 360°, so the easiest thing to do is to set your scan range to 360°, set your points to 1,050, and then change from # of points to step width, now change your scan range to 380°, and the # of points will change for you. I would always collect more data points than the bare minimum due to outlier reduction.

As for the speed, given the tolerance, you are going to be pretty slow. The book recommends 2mm/s on an Passive Head (XXT), and 3mm/s on an Active Head (Vast XT, Vast XTR, Vast). The speeds are just recommendations though. Please do your own testing.

I hope this helps get you going.

Thanks.

Edit: I have updated this post to reflect that the given parameters are from the Zeiss cookbook, and are only meant as a recommendation, not a definite on how the part should be measured or evaluated.

[[wi...]]

Please sign in to view this quote.

Thanks for the reply. I am trying this currently. I took over the quality department about 2 years ago and I have to redo everything and make a quality structure from scratch. We purchased this CMM in 2015 and our manufacturing engineer at the time setup everything and copied what our sister company in Switzerland did. No one ever questioned anything. We then sent (1) guy for basic calypso training and that was that. Everything else I've had to figure out on my own and go from there. I've learned a ton from this forum but still have a long way to go. I may have to take some Zeiss classes so I can further my knowledge.

[[Ri...]]

I've been in that same boat myself. It's difficult, and there are so many little things that you miss out of because you didn't go through training.

I remember having to call into the hotline, when I was thrown to the wolves, to ask how to calibrate a small probe (to keep it from shanking). I felt like an idiot.

We all start somewhere, and I think that for the most part this community does a good job of helping people.

[[Er...]]

Please sign in to view this quote.

Should is a strong word here. I would express my self like this:
- Filter should be 150 UPR
This number is related to manufacturing method, material and size of feature. It also dependent on if you are loocking for something special. Eg, roundness. Then you can decrease the number, to get rid of the short waves. Or if you are looking for machine marks, then you might wish to remove the long waves and keep the short. Is this a process check, or final inspection? Generally speaking, you remove more noice in the production. You tend to aim for stable results. And in the lab, many tend to look for the last micron. There is no should here.. In some cases it's even better to go for a little larger probe diameter and ease off the filters. A very hard subject to tell someone how to..
- Outlier
Do you have any defects? Is the part clean? Do you have a suitable measuring speed? (Usually lower then one might think.) That is the deciding facor here.
-Adjacent points
Why? Because the crockedbook says so? If the part is clean, and you don't have any defects. Only remove the outliers. (If any, and if you have. Why do you have them? Caused by the part or your stategy?)
Pre-filter
One should be careful using pre-filter on a standard basis. Many times this removes valid data. The only time I personally like it, is when evaluating 3d-scanned stl data in Calypso. (But that's me, a personal opinion.)

I do not wish to neglect your answer Richard. But this is a topic where it's hard to give answers without tinkering with the actual data your self. And should is a very strong word. 🤣

[[Ch...]]

Please sign in to view this quote.

My philosophy is to always know what the raw data is doing first before you apply your filters. Then, only after that, try to understand exactly what the filter/OLE is doing.

It's not just us being picky. The topic of smoothing in data science is a ginormous rabbit hole.

[[Ri...]]

I like to point people to the Cookbook because then I have at least a reference instead of just saying figure it out yourself. Lol.

Filter and outlier elimination is the biggest grey area when it comes to measurement.

[[Me...]]

Please sign in to view this quote.

For Eric,

Zeiss wouldn't have gone through all the trouble of collecting that data and giving us those *MINIMUM* specifications...It specifically states in the cookbook:

"All recipes have been proven by ZEISS as being a "good default", but depending on your special functional needs and special production processes they might not be the appropriate measuring strategies. So it is of much importance to control always the measuring results on plausibility and conformity YOURSELF."

Richard is just showing the default strategies that were suggested by ZEISS.

[[Ke...]]

As mentioned *definitely* be careful with your Filters/Outliers!

Whenever I begin a program, I *always*: open a feature ->right click in the CAD window -> select "Show Actual Points" and "Show Masked Points"

If you have a "chunk" of data that is masked from a scan that isn't located at the beginning or end of your scan, you may be filtering out more than you want to.

[[Er...]]

I'm not sure I understand what you are saying. What do you mean with "*MINIMUM* specification"? That, to me a strange choice of word in this context.

Richard (unfortunately?) said that the filter should be. Not that Zeiss recommends that as a starting point. I was trying to give a more general applicable answer. As in: Your part, and your purpouse decides what (if any at all) filter should be applied. Same as Zeiss apperently says them self, according to you post: ..depending on your special functional needs and special production processes they might not be the appropriate measuring strategies.

How many times have you asked someone what the hell they've done? And the answer you get is: "I did what the cookbook said". But hey have no clue what they've actually done, nada. I think that is a real problem, caused by the cookbook. It gives people a false feeling of confidence. Which in turn leads to less, or none at all, analyze/questioning on the result.

But again, Im not sure what you are trying to say?

[[Me...]]

Oh shoot, I'm an idiot and didn't see that nice little note you left that said "should tinker with the data yourself". I've followed the cookbook for about 5 years now and other than tweaking the speeds, I haven't had a lot of non-plausible data due to the recommended filter and outlier defaults. I've been apart of a lot of SPC-CPK & Gage R&R studies at 3 different companies now and if it wasn't for the cookbook, I would've been lost previously without constantly emailing my trainers. I was extremely angry when I found out there WAS a cookbook 5 years ago. I had already been programming for almost 10 years at that time.

I apologize Sir

¯\_(ツ)_/¯

[[Ch...]]

Whew, you can sure tell that cortizol is higher in the morning.

I'm not saying the cookbook is bad or anything. My problem is that is you give people an inch with the cookbook then they take a mile. People will be calling out people strategies with nothing more than "well that's not what the cookbook says" as if the cookbook is god or something.

[[Me...]]

Please sign in to view this quote.

You should meet the guys who trained me haha

[[Ri...]]

I haven't had enough RedBull yet...

Eric is correct, I have updated my post to reflect his comments.

Chad, I agree. I have a love/hate relationship with the cookbook. It is great for beginners who have no idea where to start, and it is definitely a lot better than just turning on the default filter/outlier and hoping for the best. Where I do have issues with it is that a lot of people end up just relying on it no matter what, and forgetting that it should just be used as a stepping stone.

At the end of the day, everything is just a recommendation, and all that matters is what you and your customer agree upon.

[[Ch...]]

I would say that for 95% of things the cookbook is spot on, or at least statistically equivalent. I'm just saying don't be a Neville Chamberlain and take their word for it.

[[Er...]]

Well, I think you simply missunderstood what I was trying to say.

And its great that you have good repetabillity, but that doesnt automatically make it the correct result. Again, it comes down to what you are looking for. High repetabillity? Well filter the shit out of everything. That last micron? Be sure its noice you remove, and not machine marks.

Suspecting a specific issue in you process, Well then band-filter can be just what you need. Not the cookbook 150upr, what you are looking for might occur around 200-300. Maybe a fourier can help you identify the span?

I guess I some how try to say "manuals" telling people what to do, instead of explaining what they do. Makes people nincompoop.

Again, I didnt try to tell "Oh Richard, your wrong bla bla bla" I just think he was in a rush and might have chosen less appropriate word. I also dont think he got offended. He has a beard, he can take it.

I would say, 5% the cookbook might be spot on. (I never know when I am spot on btw) But I guess its different from company to company, part to part 🙂

[[wi...]]

Update. I got this running smooth last week. There are slight differences between the scanning method and our typical 5 pt method which I knew was going to happen. With scanning I can actually see a roundness error not caught in our normal method. This opens up pandora's box a bit. I now get to do more testing which should be fun.

Is there a way to get a graphic element of the entire cylinder like that of a formscan or roundness tester? It would be a nice add in to see where the defect is originating from.

Also I typed this twice and had to log in? Does anyone else have this issue? I've literally disregarded some of my post replies because of the length and time I spent on a reply only for it to be erased.

[[Ri...]]

Measure it as a Cylinder, throw it into a Cylindricity characteristic, then with the characteristic selected, go to CAD - CAD Evaluation, and turn it on. You will be able to see the deviations overlayed onto the model.

[[Ow...]]

I don't have any issues with log-on using google chrome but, it may be the way IT has your firewall/internet securities set-up.

I often compare filters to practicing medicine because it’s an attempt to diagnose a part or treat an abnormality and if prescribed incorrectly, the unknown side effects can be severe. 🙄

That’s all I got to say other than that I didn’t know nincompoop was a universal word. Probably doesn’t look good on me or my siblings behalf but, my father used to use that word a lot. 🤣

[[wi...]]

Please sign in to view this quote.

I tried this. When I go to "plot" it shows up for a split second then I get a report window that states clip art failed or something of that sort?