Thread: Reliably measuring small internal features (+/_.0002") IN MACHINE (cost efficient)

The main problem in using the very machine to measure distances is that such measurements are affected by part of the inaccuracies that got the part out of specs in the first place.
If you could find a toolmaker microscope with a field of view and tenth reticle that cover the whole feature at once, fine. Otherwise, I'm afraid you'd be out of luck (as far as I know there aren't too many toolmaker microscopes with a traveling head with sub-tenth adjustments).

Paolo

There are some high end molding compounds available that may fit the bill. Another thought is Fowler Z-cat. It’s a portable DCC CMM that you could place in the machine to measure the part. I’m pretty sure this would be close to the limits of that machine but I would guess you could work up a measuring protocol that would work. One catch, you will be about 200 sets of “tenths pins” into it (>$50k) but it is multi purpose. No affiliation, just saw a demo when we were CMM shopping.

Originally Posted by Mike1974

Imagine this - first part has two features we want to inspect, (from a real print) .0684 and .0615 so there are two sets of tenth pins at roughly $230 each. We make 4 of these part. Next part has dims of (from a real print again) .0750 and .0841, two more sets of pins at $230 each. Make 4-6 of those and so on. Eventually we would have enough pins to cover them all, but still....

Check the pricing on the Deltronic site. You are actually looking at $125 a set in those sizes. If you start with a good organization system and take care of them it wouldn't take long to accumulate a useful collection.

Still not cheap but not $230.

Teryk


Sent from my XT1710-02 using Tapatalk

How accurate and repeatable is your machine? If it's tight and repeatable, you could look into probing in machine.

Originally Posted by Mike1974

I know tenth pins would solve it. End of story. However, since these are all 'customs' we would end up with literally hundreds of sets!

Imagine this - first part has two features we want to inspect, (from a real print) .0684 and .0615 so there are two sets of tenth pins at roughly $230 each. We make 4 of these part. Next part has dims of (from a real print again) .0750 and .0841, two more sets of pins at $230 each. Make 4-6 of those and so on. Eventually we would have enough pins to cover them all, but still...

I understand sometimes that is the cost of business... I am just trying to figure out a better/easier/cheaper (I know what I am asking btw haha) way.

1) I pay $124.40 per set (Deltronic +/-.0012" sets) in your size ranges from Western Tool.
2) Tight tolerances are expensive. Your customer should be the one paying for something like this, not you.
3) That said, and not knowing the rest of the part, it does seem just scrapping the first part is still the best option.

Thanks for the heads up on pricing for Deltronic pins. I guess that means McMaster is marking them up almost 100% !!

McMaster-Carr

Originally Posted by [email protected]

1) I pay $124.40 per set (Deltronic +/-.0012" sets) in your size ranges from Western Tool.
2) Tight tolerances are expensive. Your customer should be the one paying for something like this, not you.
3) That said, and not knowing the rest of the part, it does seem just scrapping the first part is still the best option.

Can't argue that, but I'm sure it would come as a bit of a shock if the prices jumped $150 overnight (I know it happens). I don't know how much these parts sell for FWIW.

I think that is going to be what we continue doing (these are small chunks of alum so mat'l isn't an issue) more or less.

Can the feature be cut undersize intentionally, taken out, measured, and then reclamped, recut?
I face this in one or small runs if the feature can't easily be measured in the machine.
Option one, make a sacrificial part but sometimes the stock cost is high or you just plain can't make a scrap one as there is only one.
So then I go to option two which is grind it with stock left to reclamp and finish and sneak up on it. That is time consuming so added costs there.

Grinding your own gage pins a option but this is an easily a 30 minute start to finish op on the timeclock so 30-40 or more dollars if you have prepped stock so again part cost.

Some of the error is machine and some is the cutting tool. I've seen it done where you take a test cut with the machine on a raw piece of junk stock to verify tool dia and machine positioning under load.

I wonder if a clamp in camera would work but as noted it will not know machine errors and is not fast.... time is money.
One can make or buy specialty micrometers that may be able to touch this...getting them to agree with your machine vision, maybe a problem.
It would be so nice if everything cut like it does in the computer and all these errors did not stack up.
One has to work with the machine is bad, the cutting tool is bad, the tool presetter is bad, the final gauge who is king is questionable.
Bob

Use a Renishaw probe. Calibrate it frequently on a ring gauge, and you can write automated probing programs that will adjust offsets as needed.

Originally Posted by CarbideBob

Can the feature be cut undersize intentionally, taken out, measured, and then reclamped, recut?
I face this in one or small runs if the feature can't easily be measured in the machine.
Option one, make a sacrificial part but sometimes the stock cost is high or you just plain can't make a scrap one as there is only one.
So then I go to option two which is grind it with stock left to reclamp and finish and sneak up on it. That is time consuming so added costs there.

Grinding your own gage pins a option but this is an easily a 30 minute start to finish op on the timeclock so 30-40 or more dollars if you have prepped stock so again part cost.

Some of the error is machine and some is the cutting tool. I've seen it done where you take a test cut with the machine on a raw piece of junk stock to verify tool dia and machine positioning under load.

I wonder if a clamp in camera would work but as noted it will not know machine errors and is not fast.... time is money.
One can make or buy specialty micrometers that may be able to touch this...getting them to agree with your machine vision, maybe a problem.
It would be so nice if everything cut like it does in the computer and all these errors did not stack up.
One has to work with the machine is bad, the cutting tool is bad, the tool presetter is bad, the final gauge who is king is questionable.
Bob

We briefly (very) toyed with that idea of cutting undersize then re-cutting, but that would take as much time IMO as just cutting and adjusting offsets and loading a new part. Op1 is always (99.9%) the side that needs to be dialed in for a couple reasons.

1) we completely profile the part, and put in (blind) dowels on this same side to avoid mismatch on the mating part
2) the opposite side (op2) usually has the center removed, think 2 'legs' or feet/standoff etc, so doing this first is not a great option

Also, if I hadn't mentioned, all this finish work is done with a 1/32" endmill. If we cut undersize and reloaded we would have to account for tool deflection too, which isn't a deal breaker I know, but in unskilled hands we could be chasing our tail. Plus sometimes the guy running these is really a parts loader...

Originally Posted by mhajicek

I'm surprised no one has mention using a Renishaw probe. Calibrate it frequently on a ring gauge, and you can write automated probing programs that will adjust offsets as needed.

Probing was mentioned I think. We would need a probe tip of about 1mm, maybe less so that would be problematic IMO. We only have a 20 tool carousel with T20 as our probe (6mm tip) so we would have to change out tips and re-calibrate/indicate runout every time, or buy another complete unit $$.

So far we are purchasing .0005" gage pins (.0115-.2505 in .001" increments) to use with the 'regular' gage pins(.011-.250) which hopefully solves most of the problem.

A place where I used to work bought +0/-.0002 pins, +.0002/-0 and half thou +0/-.0002 pins. We would mic each pin to see it "true diameter" and uses those like a sparse set of deltronic pins for smaller tolerances.

It was real shade tree but, that was what we were told to do and never had parts sent back due to non conformance (on those features).

Originally Posted by Paolo_MD

The main problem in using the very machine to measure distances is that such measurements are affected by part of the inaccuracies that got the part out of specs in the first place.
If you could find a toolmaker microscope with a field of view and tenth reticle that cover the whole feature at once, fine. Otherwise, I'm afraid you'd be out of luck (as far as I know there aren't too many toolmaker microscopes with a traveling head with sub-tenth adjustments).

Paolo

That's why we get our machine calibrated once a year. We're not talking final inspection, we're talking about making sure you make good parts.

Originally Posted by kenton

A place where I used to work bought +0/-.0002 pins, +.0002/-0 and half thou +0/-.0002 pins. We would mic each pin to see it "true diameter" and uses those like a sparse set of deltronic pins for smaller tolerances.

It was real shade tree but, that was what we were told to do and never had parts sent back due to non conformance (on those features).

I don't see anything 'shade tree' about that...

Now, if you were making the pins in-house, and claiming you had certs or whatnot maybe...

The only thing about doing it that way, is you don't get a real size, just "this pin goes, this one doesn't" which would be fine in our application so a moot point. I should also add, our cutomer is not actually checking these parts, they are part of an assembly, but we need them to work in-house. So 'technically' the size isn't the issue, it is the final result, but we get there by being able to say "the part is to print/-.0005"/+.0005" from print, etc.

As you have a vision system and the features aren't silly deep I wonder if there is any mileage in re-inventing the simple lever indicator. Basically a pivoting arm. Short at the gauging end, long at the reading end. Needs nice pivots and teensy gauging arm less than 1 mm will need careful handling.

Way back I faked up something quick and dirty with a tenths reading micrometer pushing the whole unit from side to side with a mark to set, zero, straight on, reading for the long, pointer end. Worked well enough for something better than ± 0.0005". How much better I know not but it got the job of sorting narrow parts and narrow slots into mutually not too tight, not too loose fitting pairs.

I imagine that with decent vision system you could use a bit of 1 mm scale ruler and do some sort of direct "measurement". Realistically anything better than a 0.0003" band is in the lap of the gods.

Major, major issue will be not bending the short arm.

Clive

Originally Posted by Mike1974

.
Probing was mentioned I think. We would need a probe tip of about 1mm, maybe less so that would be problematic IMO. We only have a 20 tool carousel with T20 as our probe (6mm tip) so we would have to change out tips and re-calibrate/indicate runout every time, or buy another complete unit $$.

Do you need that 6mm tip for length or something on all it's other work?
Could you just use a ruby/carbide in 1-1.5mm for everything and not have to change tips?
I know, trusting machine to be a CMM and that can lead to big problems at a tenth. 5 tenths not so bad.

Have you used handheld 1/10th pins or blocks to get a number? Does your nifty machine vision agree with the pins to the same tenth?
How is that section working out? Getting gage agreement down here is never easy. We fight gauges, check it here and you get this, check it this way and you get something else.
A probe CMM may not like your vision systems numbers. Fitted blocks or pins the same.

You started at 2 tenths and then went to hitting 5 tenths.

Since it has to "fit" go/no go is useful but you need the offline gauge to make adjustments so still high risk of scrapping part number one.

If part one does not fit your pin you have to rerun it which as you noted is a tool loading problem. We call this rerun backoff which gets complicated to keep track of.
My grinders sort of keep track of this and learn it for a part. If again needed rerun #2 adjust is so way different than rerun #1. Then back to raw stock on the next part.
The software is complex and often you look at where it will go and think that it has lost it's mind as the rerun adjust can be huge to the part tolerance.

I sort of like having multiple types of gauges and trying to make them agree within a reasonable amount but it can be very frustrating so easier if one the master.
What is the bottom line your machine can do in adjustments in the code? No machine holds one count or min increment. Not the way servo systems work.

You started at 2 tenths and now +/-5 tenths total, things change a lot here.
Bob

Mike
you could call the gage manufacture like Myers or deltronics and ask them for gage pins In increments of.0002 or .0003 or just the high and low and mean for each dim.
That would save you a ton of money rather than buying a complete set.
if your part is size .031 + or - .0003 you only need 3 pins ie .0307 .031 and a .0313(but I would get a -.0313) you dont need everything in between.

and pins are the only way to check a bore that small and that tight of a tol.

from the posts its hard to tell exactly what your doing as far as the sizes you need. the other thing your going to have to watch for is the burr on the edge. that will throw you off on very small holes. I am sure you know this but make sure you take the hole to size as the last step, dont go back and break the edge after you took the hole to size.

I wouldnt go the cheap route on pin gages either reason being is you really need that perfect smooth finish and round edge like Deltronics to check a hold accurately. pins with sharp edges tend to stick more going into the hole. Another reason NOT to make your own Tenth pins unless you can Lap a pin in and rad the edge.

it doesnt sound like your doing this in alum but if you are to keep your pins clean drop them in a ultrasonic cleaner with alchol after use, as alum grit will build up on even deltronic gages pins if there used alot for example if we have to check a ton of parts we just keep the cleaner running and lay the tool in there when were milling. we do this with thread gages alot as we 100% the threads and alcohol also works as a lubricant of sorts.

Good luck sounds like a fun but frustration job.

Originally Posted by CarbideBob

Do you need that 6mm tip for length or something on all it's other work?
Could you just use a ruby/carbide in 1-1.5mm for everything and not have to change tips?
I know, trusting machine to be a CMM and that can lead to big problems at a tenth. 5 tenths not so bad.

Have you used handheld 1/10th pins or blocks to get a number? Does your nifty machine vision agree with the pins to the same tenth?
How is that section working out? Getting gage agreement down here is never easy. We fight gauges, check it here and you get this, check it this way and you get something else.
A probe CMM may not like your vision systems numbers. Fitted blocks or pins the same.

You started at 2 tenths and then went to hitting 5 tenths.

Since it has to "fit" go/no go is useful but you need the offline gauge to make adjustments so still high risk of scrapping part number one.

If part one does not fit your pin you have to rerun it which as you noted is a tool loading problem. We call this rerun backoff which gets complicated to keep track of.
My grinders sort of keep track of this and learn it for a part. If again needed rerun #2 adjust is so way different than rerun #1. Then back to raw stock on the next part.
The software is complex and often you look at where it will go and think that it has lost it's mind as the rerun adjust can be huge to the part tolerance.

I sort of like having multiple types of gauges and trying to make them agree within a reasonable amount but it can be very frustrating so easier if one the master.
What is the bottom line your machine can do in adjustments in the code? No machine holds one count or min increment. Not the way servo systems work.

You started at 2 tenths and now +/-5 tenths total, things change a lot here.
Bob

I don't think I changed what I said we would like to do, I said what we are going to do to work towards this problem. The half thou pins will increase our accuracy by 50% since now we can only check with .001" increment pins, factoring in the pin tolerance of +0/-.0002 of course. I personally think I can guess pretty well how a pin is fitting ie- I had to kind of push it to fit, it drops in easy, but the next size up won't start,etc, but that is me and I am not the one checking these every single time.

And no, the vision does not always agree with a hard gage (of any kind), we.. I guess for lack of better terms, 'split the difference'. Also, because the vision system relies on lighting and operater skill, we would probably go with a hard gage if there was a 'fight' over it.

I have used tenth pins, but not at this place.

I don't know if we could keep a 1mm tip fitted all the time. I think it would cause problems as sometimes we do need some length for probing.

Originally Posted by Mike1974

I don't know if we could keep a 1mm tip fitted all the time. I think it would cause problems as sometimes we do need some length for probing.

Can the job support the purchase of a second probe dedicated to this purpose?

You mentioned Haas earlier, so I guess that means no glass scales, no thermal mitigation etc. No concerns there?

I guess if the feature is parallel to one axis and over a short distance it should be very close in any case, as long as you don't care about absolute position, and only comparative measurement.

2 tenths, you are probably going to need to consider lobing error, either by orienting the probe as necessary or maybe a strain gauge probe might be necessary?

Buy gauge pins .0100 to .0110 by tenths then use gauge block of suitable size and use pins fo the difference or feeler gauges if you can get a good set.