Spc ???

We haven't done an actual charted SPC here in more than one coons age, and even then I don't recall what we did or would have done.

Of course you check and doc your findings bla bla bla. No problems there...

BUT - what happens when you find out of tol parts, or you have a train wreck, or ????

You check your part and it's out of tol. You shut down, adjust/retool/whatever and fire back up aggin. Then you go and sort all the offending parts out of the local tray.

How is this shown?

I mean - you hafta show the out of tol parts check eh? BUT - those parts aren't making it out the door. You are going to toss the one(s) that you checked as well as go sort for more.


-------------

Think Snow Eh!
Ox

I don't think SPC is worth much to a shop that runs small lots or just jobs.
If the customer requires it, do it, but charge them for it, all of it.
The last larger shop I worked at bought in, and got hosed, it generated several suitbilly jobs and spent a lot of money to little effect.
I see this in the medical field I work in too. A journal article has all kinds of statistical analysis done on cohorts numbering ten to twenty patients-worthless, but it looks good to those who don't know the math.

The correct way is to chart it and assign it a special cause and apply apropriate corrective actions. Then when it is questioned you then revert to your nonconforming documentation.


I agree SPC is not useful for small runs, however 30 pc capability studies are as they are the foundation for estabilishing SPC programs. An even quick and dirtier way acheive a level of statistical process approach is to do pre control where you establish process control limits at I think 66.6% of your tolerance. Any readings beyond that would need further evaluation.

Fun with numbers

I need out of quality and be value added

first there was quality control, tag rejects and keep going
then quality assurance, tag rejects and try to repair tooling
then SPC, statistical process control, record product state with data and dimensions which change over time producing a maintenance guide line bell curve indicating preaction to problems.

Yeah, the real question is why you would do this to start with. If a customer requires it, get a good reference book, learn the lingo, and discuss with the customer what is expected - and why they want to see this sort of program. If this is for your own internal use, the key is that you know from the documents you create what you did, and why you did it. As others have noted, trying to apply SPC to small qty parts runs is nuts from a practical standpoint.

countryboy1966 said:

I agree SPC is not useful for small runs, however 30 pc capability studies are as they are the foundation for estabilishing SPC programs. An even quick and dirtier way acheive a level of statistical process approach is to do pre control where you establish process control limits at I think 66.6% of your tolerance. Any readings beyond that would need further evaluation.

Click to expand...

For precision machining - where tool wear is the primary variation - X hi/lo-R charts with control limits set at 75% of the tolerance is the best recommendation. The problem is a sampling error. For usual Shewhart charts, only one measurement per part is taken. Think about a circular feature. How many diameters are in a circle? An infinite number. How many points are you taking? One. One out of infinity is a statistically insignificant sample, and does not capture the important within-variation. X hi/lo-R technique captures all of the diameters, and collects "within" variation, which ends up representing roundness. This charting is much more meaningful for precision machining, and the control limits are not generated by the wrong distribution - which helps a lot! Any points beyond 75% means it is time for re-adjustment for tool wear. The result should be a sawtooth curve, not random variation about a mean. If you have random variation you have one of two likely issues: 1) overcontrol or 2) measurement error.

specfab said:

Yeah, the real question is why you would do this to start with. If a customer requires it, get a good reference book, learn the lingo, and discuss with the customer what is expected - and why they want to see this sort of program. If this is for your own internal use, the key is that you know from the documents you create what you did, and why you did it. As others have noted, trying to apply SPC to small qty parts runs is nuts from a practical standpoint.

Click to expand...

Actually, there is a good statistical reason why for most short runs in machining SPC is of virtually no value. Warm-up and start-up are special causes (that can not be eliminated economically), where there is known instability. It is not until you get past these points where any SPC becomes valid. Not only that, for precision machining, you need to use the correct SPC technique. X-bar/R is the worst technique - bar none - for SPC of precision machining. It encourages overcontrol, and only plots variation caused by measurement error. Not only that, the X-bar/R (and I-MR and any other Shewhart charts) uses incorrect statistics for precision machining. For machining that is primarily affected by tool wear, the X hi/lo-R charting of lengths and diameters is far more powerful, provides more information for process control and uses the correct statistical distribution.

Bob Doering said:

Actually, there is a good statistical reason why for most short runs in machining SPC is of virtually no value. Warm-up and start-up are special causes (that can not be eliminated economically), where there is known instability. It is not until you get past these points where any SPC becomes valid.

Click to expand...

So for a small shop with 5-10 part runs SPC is worthless as you can't eliminate the special causes?
If all your runs are small won't the "special causes" (across many runs) form a normal distribution allowing you to predict your process capabilities?

Even if your distribution is flat and hard limited (parts outside are thrown away) your averages (X-Bar) will still show normal won't they?

Yes, I agree, it takes a lot more thinking on how to handle the data in a small run shop and this falls outside of most basic SPC training.

The biggest point I always make is you should never, never, ever do this for your customer.
You do it to pump more good parts out the door as fast as possible.

If it does not help you it is a waste of time and money and is not working.
Don't learn the lingo, learn how to make work for you.
You don't buy a special boring tool or machine just because others have one, You buy one to make money.

If it doesn't put more money in the checkbook it is a useless tool.
Bob

CorrectSPC - SPC for Precision Machining - Stop the X-bar R Madness! - YouTube

CarbideBob said:

If all your runs are small won't the "special causes" (across many runs) form a normal distribution allowing you to predict your process capabilities?

Click to expand...

1) The point of SPC is to seek out special causes and react to them. here, you would be looking for special causes within special causes.
2) A normal distribution of data from a known special cause has no particular value. Shewhart charts do not rely on a normal distribution, but they do rely on independent, random variation. That is not what tool wear is, and that is why X-bar R charts are statistically incorrect for precision machining.
3) Standard Shewhart SPC charts (x-bar/R, I-MR) have no relation to capability.

CarbideBob said:

Even if your distribution is flat and hard limited (parts outside are thrown away) your averages (X-Bar) will still show normal won't they?

Click to expand...

1) Again, you already know you have the special cause of warm-up and set-up. Even if the distribution is “normal”, any further special cause will be masked by warm-up and start-up.
2) For precision machining, a normal distribution is a sign that the process is not in control! It shows either overcontrol , excessive gage or measurement error (two different things) or instability. The charting should show a steady trend from tool wear, with adjustment, generating a sawtooth curve, which is the continuous uniform distribution. In precision machining, normal is bad.

CarbideBob said:

Yes, I agree, it takes a lot more thinking on how to handle the data in a small run shop and this falls outside of most basic SPC training.

Click to expand...

Absolutely. I recommend using X hi/lo R charting – even for short runs –as it will tell you when to adjust and will assure you that no bad parts will be made (unless the tool breaks - special cause - then it will tell you bad parts have been made.) The key is that you do not use averages – which hide your within-part variation, as well masking important clues to your product quality. For precision machining, X-bar /R is worthless – and that does fall outside of basic SPC training (and most trainer’s expertise).

CarbideBob said:

The biggest point I always make is you should never, never, ever do this for your customer.

Click to expand...

Some customers want assurance that you have a clue what you made. It is “report card charting”. It is not as powerful as process variable charting, but it does have a point. However, if it is using the wrong chart, it will confuse both the customer and the operators, and that is dangerous and wasteful.

CarbideBob said:

If it does not help you it is a waste of time and money and is not working.

Click to expand...

I agree!!! That is why you should not use X-bar R in precision machining! Use something that actually plots the GD&T of the feature!

CarbideBob said:

Don't learn the lingo, learn how to make work for you.

Click to expand...

Don't learn the lingo, learn how to use it correctly! If it doesn't work for you, you are not using correctly. If you have to make up "special rules" - especially for control limits - it is a sign it is either not working or the wrong tool. You can not "make it" work for you. But, the right tool, implemented correctly, will work for you.

CarbideBob said:

If it doesn't put more money in the checkbook it is a useless tool.

Click to expand...

I agree! That is why X hi/lo-R charting is so valuable. It gives you correct insight when to adjust your procss, can tell you your tool wear rate per part (and, therefore the effectiveness of your tool), can show you the effects over time of the tool pressures between roughing and finishing tools, can even tell you when to change your tools before they break, in some cases! X-bar R charting can not do that!! So, use that tool that can put money in your pocket - and takes no more (if not less) effort than those old X-bar R charts!

In a nutshell:
• The X bar chart from the X bar – R charts represent the average of a statistically insignificant sample of measurements for a of a circular or linear feature
• R charts from the X bar – R charts represent the range of measurement error

Neither of those charts, therefore, provides any data that accurately represent your process!

• Control limits are calculated using statistics for the wrong distribution – the normal distribution. The correct distribution for precision machining is the continuous uniform distribution.

• X hi/lo – R charts represent the GD&T characteristics of a circular feature: diameter and the zone represented by roundness (or length and parallelism for a linear dimension)
• X hi/lo – R charts provide more valuable data, such as tool wear rate
• X hi/lo – R charts use the correct uniform distribution for precision machining - the continuous uniform distribution
• X hi/lo – R charting techniques can be expanded to control taper

Remember this:
• Automated tool wear compensation is not statistical process control. It is codified overcontrol.
• The algorithm for compensation becomes the process, not tool wear
• Lose some benefits of SPC because the constant adjustment masks the information
• Would help if the compensation was tracked

If you are interested in free, quick video training on the correct methodology for SPC for precision machining, click here

Well I really hate to hijack Ox's thread here..... (Ok that's not very truthful, I love pokin at him.)
He nailed the best possible title so this seems like a good place.

I'm sure it is not a surprise to most that I am a SPC "believer".

Has anyone else sat through this link?
What was your gut reaction? Helpful? Heard this from the upper levels too many times?
More BS from the QC guys? Ammo that you can use to nail the boys above with?

Bob Doering said:

CorrectSPC - SPC for Precision Machining - Stop the X-bar R Madness! - YouTube

Click to expand...

I'm kinda hoping we can get a lively discussion going on this whole SPC thing that has been shoved down onto too many suppliers and shop floors.

I know my viewpoint is tilted, much like the street drug dealer and pimp who has now found salvation, or worse yet the 2 pack a day smoker who has now quit and can't stand the smell.

Does it work? Can it be used in the small run shop?
Is it just a pain in the ass invented by pencil pushers?
An absurd concept used to push the smaller guy out of the loop?
The dumbest thing ever invented?
Does any of this buzzword loaded system make any sense?
Is this all just "common sense" with a worthless paper trail attached?

Maybe I'm just wasting time, but as a builder of gauging and SPC systems for 25+ years I'd like to hear from the field.

Most of what I see on the net is absolute bullchit on this subject. (my humble opinion here, and those that have read my posts realize that I know very little)
Experts with no chips in their shoes (talk to me when you have a million or so parts that you have actually run all by yourself) or people who get a system jammed onto them to make the QC people happy with a stupid form that just gets in the way of making parts.

A note to Bob: the author of the video, get your flameproof suit on, I would expect some heat here, this is not an easy crowd.

Don't be bashful (as if this is a problem here). Let's hear your feelings on the subject.
Come on guys, give me some feedback that we can argue (..oops,, um, debate) about.

Hoping this will light a few fires.
Maybe we can get this section of the forum down into some real shittin and gittin.
Bob

CarbideBob said:

A note to Bob: the author of the video, get your flameproof suit on, I would expect some heat here, this is not an easy crowd.

Click to expand...

No problem. I have run this past some tough audiences- including Japanese customers, Master Black belts, union operators, statistics gurus and ASQ sections around the country. If you sit through it, you will realize this comes from the frustration of trying to do X bar-R charting in precision machining, and it did not work. That is what gives SPC a bad name in the machine shop! The examples in the video are real machining issues that were resolved. This technique - and the resulting chart - should actually make sense to people that run machines. It sure did in the shops that I implemented it in.

But...before people throw gas on it and light it, they ought to have the guts to try it first. Hip shots really don't cut it...that is the real B.S. I put it on here for free to help people out. I hope they get that point.

CarbideBob said:

....or people who get a system jammed onto them to make the QC people happy with a stupid form that just gets in the way of making parts.

Click to expand...

An interesting side effect of the X hi/lo - R chart is that it cuts back on the number of adjustments operators tend to make as they "run to the mean" (classic overcontrolling) when they don't chart. Yes...less work, less messing with the machine. Can't say that is "getting in the way of making parts"! Instead of pushing buttons on the machine, they can use the time to measure a part or two..and push buttons on a calculator. Let the machine do its job.

Ox said:

BUT - what happens when you find out of tol parts, or you have a train wreck, or ????

You check your part and it's out of tol. You shut down, adjust/retool/whatever and fire back up again. Then you go and sort all the offending parts out of the local tray.

How is this shown?

Click to expand...

The point of correctly implementing SPC is to give you the clues to make the adjustment or change the tool before you make the bad parts! Saves all that wasted material, time and sorting - plus saves you from the part you missed when you tried to sort.

It will not help you as much in "special cause" train wrecks - such as unexpected tool breaks, power outages, etc. But, if it does the above you are still ahead of the game.

True, X bar-R charts will not give you much info to guide your decisions, but the X hi/lo -R chart will (as long as your gage is of proper resolution).

Charting is a like a gas gage. It is true - to support many people's claim about charting - you can drive a car without a gas gage (for years and years, if you chose). But, would you really want to? What is the down side of actually knowing what your process is doing?

specfab said:

Yeah, the real question is why you would do this to start with. If a customer requires it, get a good reference book, learn the lingo, and discuss with the customer what is expected - and why they want to see this sort of program.

Click to expand...

Problem is, most books are wrong when it comes to SPC for precision machining (especially short run) - so this approach is likely to confuse both you and the customer, make useless work and cause lot of bickering. Even the 30 pc capability study is typically performed incorrectly, and the result is only valid in machining if your lot size is...well, 30 pcs. Better idea - watch the video link above. It explains the whole capability problem with precision machining.