I have a problem,

I am trying to calculate the upper and lower control limits for a hole position with a size as follows:
1mm diameter +0.10. Please note this is a geometric tolerance and as a result equates to nominal plus 0.10. Standard control limit rules only allow for linear values. A hole position can be out of position through 360'. Similar to the appearance of a scatter graph.

Control Limits I think do not allow for this? Or do they?

The way I'm thinking is that you have a value of say. . 10.00+/-0.1. The values can only fall from 9.9 to 10.1 in a linear manner. Whilst a hole position can fall anywhere around a single point not just above, below, left and right? Or am I making no sense whatsoever?

Please assist

Lee Moffatt (:confused: )

Can you run SPC in both the x and y axis?

Geoff

No,

All I have is a value at the present time.

I have been investigating it further and understand that I do not need a lower cntrol limit because is there is none! There is an upper control limit obviously but the lower is in fact the target.

Therefore the cp value is reduntant but this still does not help in my quest for understanding if I can apply the ucl limit to this type of characteristic

Lee Moffatt:confused:

Lee,
You may want to check out an article from Quality Mag from February 2001 called "Simple Process Capability?". It is at qualitymag.com go to the article archives. I think this is what you want.
Russ

I received this in an e-mail:

Forgive my intrusion,

I have been reading your replies to spc & GD&T and would like to view the file as named above in the subject bar, if you still have that file.

I have posted a new thread entailed GD&T And SPC that describes the same type of problem.

I have this old file of Marty Ambrose's. It's dated 1999. I don't know what ever happened to ol' Marty, but he was one of the 'pioneers' here at the Cove in the old forums.

I believe Don Winton also wrote a paper on this which was in the old pdf_files directory at one time. I will look for it this weekend and if I find it I'll post it in this thread as well.

That said, this file is, and has been, available in the Members and Premium directories. You might want to be nice and consider donating to 'the cause' --> http://Elsmar.com/join.html (plug intended...).

The article is excellent, about hole drilling I found an article about ussing "A PROCESS CAPABILITY INDEX
SENSITIVE TO SKEWNESS" by Peter A. Wright, I tink it could complement it.

Lee,

I've got a couple of questions. What is the tolerance for feature size and what is the geometric tolerance for position? Is the position tolerance constant or variable (does it have a MMC modifier attached to the tolerance)?

I don't suppose that you are questioning about the upper and lower control limits for feature size.

Process control examination for the position tolerance (I think) is best accomplished by monitoring the X,Y,Z coordinates of the feature location separately. It provides two important benefits, It helps to distinguish whether the individual coordinates are accurate (mean centered on the basic location) or not and it helps to discover how precise each axial deviation is and how much each is contributing to process variation.

You commented about the position tolerance being non-linear scattered about 360 degrees. The position tolerance describes the zone that the axis, median plane, center, etc. of the feature must reside within. The zones can be described in a number of ways (spherical, circular, cylindrical, rectangular, square, cubic, etc.) all determined by the symbols used in (or absent from) the feature control frame and from the way the tolerance "leader lines" are depicted on the specification.

The individual coordinates can be normally distributed when examined separately but when they are combined to determine the displacement from the basic location the resultant radial separation (or doubled "diametrical deviation") is typically a skewed distribution that is 'nearer to' and 'truncated by' zero and tailed toward the USL. Monitoring the derived position deviation for process control is not good for a couple of reasons. The position deviation does not distinguish between process parameters "X, Y, & Z" that are accurate but not precise where mean-shifts are no help and ones that are precise but not accurate where mean-shifts may improve. The other reason is that the deviation is not typically a normal distribution.

The control limits for the coordinates are established by the process variability so there is no relation to the drawing specification for position. Process control can be done effectively
by monitoring the X Y & Z of geometric deviations.

To predict the process capability for the true position deviation one must first establish that the process is "in-control" and that can be done with the individual coordinates. If the geometric tolerance is constant one can predict capability by applying the appropriate distribution function that best fits the skewed true position deviations and transforming the data. If the geometric tolerance is variable you can use a method that compares the (USL plus the Mean variable tolerance "bonus tolerance" minus the Mean geometric deviation) to (three times the square root of the combined variances for size and geometric deviation). It assumes normality for both but it demonstrates prediction error margins comparable to predicting the capability of a constant tolerance with the Weibull Method.

I would not recommend using: the residual tolerance method described in "Simple Process Capability" Quality Magazine by me, the percent of tolerance method described in "Calculating MMC Cpk" by Marty Ambrose, or the adjusted true position method "Calculation of Cpk under conditions of Variable tolerances" Quality Engineering by Glen Gruner because each one of those methods uses an individual pair of variables for size and geometric deviation to produce a surrogate variable the can be compared to a constant limit. In so doing the underlying variation from the independent sources can either be amplified or moderated in the surrogate.

I will be publishing another paper soon.

Both, Upper as well as lower Control Limits will exist whether you have geometric tolerances or the traditional ones.

Geometric tolerances will define circular tolerance zones for location of a round feature, and will have a one sided tolerance. Cpk will be defined as Cpk(upper) and Cp will be undefined.

As for Control Charts, it will pay to plot separate control charts for tolerance of size and tolerance on location.

In case there is a bonus tolerance allowable, by a modifier applied to the location tolerance, the method sited above by Marty Ambrose will tell you how much of the allowable tolerance is 'eaten up' by your process and give you a very good estimate of the proess capability, after compensating for the bonus tolerance.

It will however help to plot the control charts for location, without giving any consideration to the bonus tolerance. Control Charts can be used to monitor the inherent stability of the process, without paying heed to the tolerance (bonus or otherwise).

Geometric tolerances of round features will commonly be round but not always. In the American national standard the shape of the zone is defined by the symbol preceeding the tolerance or by the way leader lines are attached to the features. See page #140 of ASME Y14.5M-1994 for an example of how the tolerance zone could be defined to allow extra rotational freedom about the secondary axis (possibly for an application such as the position of a pinion shaft in a planetary gear assembly).

As for control charts for position specifications, it will pay plot separate charts for the X, Y, and Z coordinates. To reduce the variation of a geometric position distribution one would first have to detect from the individual XYZ coordinate locations whether the position deviation can be reduced with mean-shifts to X,Y,Z or not. Since the charted variable of computed position deviation commonly includes the magnitude without its direction, the radial deviations of a group of clustered in the first quadrant of a Cartesian coordinate system at the maximum radius can have the same mean and variation as a group scattered on the perimeter at the maximum radius if the polar that direction is ignored. The computed radius or diameter of geometric deviation does not indicate whether there are potential process improvement opportunities to reduce variation via mean-shifts or not. A process that is accurate but not precise can only be improved by a reduction in common cause variation. A process that is precise but not accurate can benefit from shifting the means of the coordinates to their target values but the two distributions are indistinguishable when only the radius or diameter of true position is the charted value.

In the case where the tolerance itself is variable (by permitting bonus geometric tolerance relative to feature size) I personally don't reccomend the method sited by Marty Ambrose with what I know of the shortcomings of various methods predicting the capability of processes with variable tolerances. Attached is a response that I wrote to Todd Minnick after I examined his work with variable tolerances.

Let me say that I am indebted to all of those that have and are
attempting to find methods to predict conformance with variable tolerances. Glen, Marty, Todd and others are pioneers in this stuff.

Can you run SPC in both the x and y axis?

Geoff
Hello sir
I want to plot the control chart,range chart & calculate the cp,cpk,for flateness required within 0.1mm.pls. help me how can i draw the above becouse there no LSL.
Ranjeet Singh:confused:
Chandigarh

Welcome to the cove, I found this place enlighting and I am sure you will find it too.

plot the control chart,range chart & calculate the cp,cpk,for flateness required within 0.1mm.pls. help me how can i draw the above becouse there no LSL.
Ranjeet Singh
Chandigarh

Control Chart has nothing to do with specs, so you can draw them the standard way without any worry.

Cp can't be calculated because it takes in account both specs:frust:

Cpk can be calculated (only the upper limit <Cpku> of course) and some guys may think is a good idea, I take better Cpmk, at least it takes the target in account, if you use the Cpku as Cpk you are without knowledge taking Cpku = Cpkl, altho it doesn't exist.

Sir pls. tell me about DOE,How can we get help in process manufacturing auto parts of casting(SG,CI),forging and machining .:rolleyes:

Sir pls. tell me about DOE,How can we get help in process manufacturing auto parts of casting(SG,CI),forging and machining .

I think you may get more answers if you post your question as a new topic, because most or the people check the topic description, and DOE is a little bit out.

Well, it a too broad question, DOE can help any process as a tool to detect how variation in some specific conditions affect the product, so you may find some conditions where you may have a more robust product (to the variations in your process), have less problems with out of specs parts or make the produt behave better (with negligible or without effects on the end product) as part of a product improvement or materials reduction program.

Hello sir
I want to plot the control chart,range chart & calculate the cp,cpk,for flatness required within 0.1mm.pls. help me how can i draw the above because there no LSL.
Ranjeet Singh:confused:
Chandigarh

Actually, you do have a lower spec - 0 mm. Hard to get better than that. But, the biggest problem is that the data will not be normal. Unilateral tolerances (max flatness) with 0 as a minimum only behave normally when your process is bad. The closer it gets to 0, the curve "smashes" against the 0 limit. I believe this behavior is closer to the Weibull curve, although others may know of a better curve. Some have pretended that the distribution is half of a normal distribution, but that is a backyard estimate. I would recommend using a 75% of tolerance upper control limit. An individuals chart may make the most sense, since the variation of flatness of 3 or 5 consecutive pieces should be negligible, if you are doing precision machining. Spend the measuring time making sure you are covering the designated area as completely as possible. Set your upper range limit based on the expected range found in consecutive parts from your capability run. Anything greater than that should be considered unexpected or "special cause". Make sure you have appropriate gage R&R for that measurement, too. And, since the distribution is truly not normal, Cpk, etc., calculations do not apply (see AIAG PPAP 4th ed: 2.2.11.5 Processes with One-Sided Specifications or Non-Normal Distributions) - so do not bother with them. The key is to understand you flatness variation, and how to "control" it. Can you adjust - or "dial in" - your process for flatness? If your chart can tell you when to "adjust", it is working. If not, it may just be a report card chart,and an academic exercise at best. :cool:

Actually, you do have a lower spec - 0 mm. Hard to get better than that.

In a unilateral tolerance zero (0) is not a specification limit. It is the ultimate achievement or ultimate goal. If one had a flatness tolerance of .020, does that mean that the USL is .020 and LSL is 0 with a nominal value of .010? Absolutely not!

If one wanted to calculate capability of flatness, one could calculate the Ppk and Cpk but not the Pp or Cp since they require specification limits. Ppk/Cpk are calculated from the process average to the only specification limit which is .020 in this example.

Flatness is not an appropriate feature or characteristic to apply SPC unless one has a CMM and a scan mode but that is unlikely on the shop floor. We all measure it differently and most likely, set it up differently. We could get data of some sort but it is possible that one part may look statistically good but another person may measure it out-of-specification.

Very few, if any, GD&T symbols are appropriate for SPC applications.

In a unilateral tolerance zero (0) is not a specification limit. It is the ultimate achievement or ultimate goal. If one had a flatness tolerance of .020, does that mean that the USL is .020 and LSL is 0 with a nominal value of .010? Absolutely not!

If one wanted to calculate capability of flatness, one could calculate the Ppk and Cpk but not the Pp or Cp since they require specification limits. Ppk/Cpk are calculated from the process average to the only specification limit which is .020 in this example.

Flatness is not an appropriate feature or characteristic to apply SPC unless one has a CMM and a scan mode but that is unlikely on the shop floor. We all measure it differently and most likely, set it up differently. We could get data of some sort but it is possible that one part may look statistically good but another person may measure it out-of-specification.

Very few, if any, GD&T symbols are appropriate for SPC applications.

I agree, there would not be a nominal. Nominal is really a non-issue for non-normal distributions. Ppk/Cpk are calculations that relate to normal distributions, so they would not apply to this situation at all. (see AIAG PPAP 4th ed: 2.2.11.5 Processes with One-Sided Specifications or Non-Normal Distributions) Again, it is only normal if you are so far away from 0 that the full distribution is normal - which is typically if you have a bad process. The distribution should be skewed towards 0 if the process is of any value.

I agree, it is not as easy of a feature to control as length or diameter, but if it is truly critical to the function, it is worth generating a methodology to control it - preferably a visual one to spot special causes before a bunch of scrap is made.

As far as one person finding the feature good and another finding it bad, that is a measurement problem that needs to be resolved between the customer and the supplier in any event. An agreed upon measurement technique must be established or there will never be agreement between the parties - and indiscriminate rejections will plague the life of the part.

Very few, if any, GD&T symbols are appropriate for SPC applications.

Actually, when SPC for precision machining is done correctly, roundness (for circular features) or paralellism (for lengths) is perfect for SPC applications (as a measurement of range). But, that is a different discussion. :topic:

Very few, if any, GD&T symbols are appropriate for SPC applications.

Nonsense!

Paul F. Jackson

Nonsense!

Paul F. Jackson

Paul:

Good to see that you are back.

Nonsense!

Paul F. Jackson

Paul,

Would you be so kind and elaborate on your post?

Thank you so much.

Stijloor.

I'll jump in; any measured dimension or property can have SPC applied. there is nothign magical or mystical about GD&T. However, one must first differentiate between product acceptance and SPC. they are not the same. It is true that things like max material condition 'complicate' acceptance rules but they have no effect on SPC...the feature is what it is. control that.

The true position of a hole can be controlled in the x axis and y axis OR as a vector inthe circle. THINK about the physics of the process and not GD&T rules. Then it will become clear how to control the process.

How long have we been saying that spec limits aren't control limits; that process control is not product acceptance?

And process capability can also be quantified but one must think beyond cookie cutter AIAG 'standard rules'

<snip>The true position of a hole can be controlled in the x axis and y axis OR as a vector in the circle. THINK about the physics of the process and not GD&T rules. Then it will become clear how to control the process.

Good points. However, it is not that simple...

For example: The (automotive) customer (whether we like it or not) requires some sort of capability index on position. I have yet to find two (quality) engineers that agree on a valid method on how to do that. If you have followed some of the (heated) discussions here about GD&T call-out interpretations, we have a LOT of work to do...let alone agreeing on how the data should be generated, collected and analyzed to come up with some valid assessment for the purpose of statistical process control.

If it was that easy....:(

BTW, we must think about the GD&T rules, that's what Y14.5 requires.

Stijloor.

Well stated Bev!

To say that “Very few, if any, GD&T symbols are appropriate for SPC applications” denies the enormous benefit that predictive statistics has in comparing observed measurement variation to defined boundaries for geometry. GD&T and SPC are both very powerful tools that, when applied efficiently, can not only describe the boundaries of acceptable variation in terms of fit and function but also reduce the scrutiny required to characterize, monitor, and adjust the parameters of a process that generates that variation.

Just as there is great potential for these tools to complement each other when understood and employed correctly there is equivalent potential for them to incorrectly define or characterize acceptable variation when one or both are used inappropriately. Pioneers of the tools, Walter A. Shewart (SPC sub grouping, histogram, X-Bar and R-Bar charting) and Stanley Parker (GD&T, MMC, LMC and attribute gauging), created simplified practices that address the complex nature of the analysis tools.

Industry expectations of quality practices however, have changed from the time that the tools were first employed. Most customers now require that producers demonstrate predictive conformance to the product specifications in terms of process capability risks, namely Cp, Cpk, and/or Pp, Ppk ratios. These indices require continuous data analysis rather than the discreet data generated by the attribute gauging.

Unfortunately the simplified practices that normalize data by sub-grouping and use ranges to estimate control limits… as well as… reporting discreet data pass/fail statistics with regard to virtual condition limits… are generally abandoned in favor of raw measurement data to perform the capability analysis. In so doing the analysis becomes more complicated. Distribution types must be examined, normal or otherwise, and best fitting curve functions employed. Separate data for size and position of variable limit tolerances must be analyzed together to reflect the virtual condition boundaries that are equivalently built into attribute gages.

Typically these precautions are overlooked simply because people don’t understand the prerequisites of statistical analysis, or do not recognize that there are variable limits with many geometric tolerances. It is quality practitioners that fail to check the data for control (randomness), assume that all data is distributed normally, and disregard feature size as a parameter for geometric position tolerance. I don’t blame them however it is those should know the limitations and prerequisites of the analysis but do not… those that demand demonstration of capability (STA & Purchasing), those that govern/solicit statistical analysis procedures (AIAG & Software Manufacturers), and those that are regarded as experts in this stuff that do not speak out about the abuses! To be fair to them although, I have found that there are experts in SPC and experts in GD&T but to borrow Dave’s words “few if any” are experts in both.

Bev pointed out, as I have to Dave earlier that parameters for process control do not require specification limits… only checks for randomness and predictability, furthermore those parameters, if chosen and monitored properly, can help the process owner to identify characteristics of the process that may be improved with adjustment. A surface that is specified parallel to another has at least three components that can be monitored… its flatness and its pitch and roll! When the variation of pitch and roll of the median plane are nominally aligned to the datum feature plane then the resultant variation in parallelism represents the “entitlement” Cp or Pp… when they are not aligned the value represents measured capability Cpu or Ppu.

Stijloor, I hope this isn’t too much of an elaboration.

This stuff needs to be fixed if we are going to continue to demand Cpk’s of Geometric tolerances (with variable limits) from producers. Otherwise drop the capability requirements and go to near 100% attribute gauging as Dave suggests, to achieve required capability levels.

Paul

Hello Lee,

Attached is a presentation on GD&T. Hope it may help you.

I stated previously that “Very few, if any, GD&T symbols are appropriate for SPC applications” but I did not state that we could not use data generated from confirming GD &T symbols in some sort of statistical information if it is helpful.

Let’s take the positional tolerances of 10 holes as an example. We have a feature control frame with a diametrical tolerance zone of 0.3 beyond MMC referencing datums A, B (MMC) and C (MMC). Datums B and C are holes.

This means that each of the 10 holes relative to their true position (theoretical centres) are located from datums B & C, perpendicular to datum A, accrued tolerances from individual hole sizes on each of the 10 holes, shape of each hole impacting its tolerance (virtual condition boundary supersedes CL), datum holes at MMC possibly accruing additional tolerances. All are to be confirmed simultaneously.

It is appropriate to separate only 2 of the requirements and plot their location while disregarding the remaining ones? Is it possible that the resulting distribution may appear quite good but we have a couple of parts that are out of specification caused by, say, the angularity of a hole?

We also have a unilateral tolerance (round) that must been calculated on each hole. The true position (zero) is not a lower tolerance zone so how should one handle the plotted points? Maybe we could vector them (X & Y co-ordinates) but the tolerance zone is round? I have seen people try using a +/- 0.15 in place of a diametrical tolerance zone of 0.3 but that is plainly incorrect. There is a way of reflecting the plotted points using a tolerance proportion but complicated.

Now if one plots points to analyze the pattern and it might be helpful, great, go for it, but capability studies as well as ongoing SPC on the shop floor are not appropriate nor practical.

The only practical way to confirm the simultaneous requirements of positional tolerances is with an attribute checking fixture with pins of virtual condition size staked in each of the 10 holes (force of 1 finger) with the part mounted on datum A and datum holes B & C on pins of either MMC or virtual condition size. This is appropriate method of confirming the positional requirement and simulates the assembled condition.

Hard gages cost huge amounts of money and provide zero information, they are used for large runs and as a catch for bad parts, not to make adjustments to a process. They are pretty much junk gages. A cmm for bolt patterns will provide the numbers needed and that is what you would want to use for such a 10 bolt hole study, and ongoing long term studies. Worst case you would use a hieght stand and calc out everything. A hard gage will not tell you how to adjust to make a passable part, only that it fails.

SPC and GD&T are fine together, there is some contention about reporting the X, and Y or the callout. I report the callout and not the dims that make up the callout.

If I was doing a study on the machine I would report the X and Y, a study on the process to meet customer specs reporting only the spec callout is correct.

Now that I have thought of it it has been over 10 years since I last saw a hard gage for TP on a shop floor being used. I do find it surprising anyone uses them still. I guess there is a need if you have huge runs and no cmms.

Good points. However, it is not that simple...

For example: The (automotive) customer (whether we like it or not) requires some sort of capability index on position. I have yet to find two (quality) engineers that agree on a valid method on how to do that. If you have followed some of the (heated) discussions here about GD&T call-out interpretations, we have a LOT of work to do...let alone agreeing on how the data should be generated, collected and analyzed to come up with some valid assessment for the purpose of statistical process control.

If it was that easy....:(

BTW, we must think about the GD&T rules, that's what Y14.5 requires.

Stijloor.


My experience is that it's not all that difficult once you separate the needs of process control and product acceptance. even for capability indexes. The automotive industry does allow for alternate methods of calculation when the distribution is not normal and GD&T controlled features certainly fall into this category (the use of MMC alone does this). One can use a simple defect calculation and back calculation of a capabilty index. I've never had difficulty with this as long as I talked to the supplier rep with the data in hand...

Yes you must think about GD&T rules for product acceptance, but there is no law that says that GD&T applies for controlling a physical process...In fact if you monitor your process, improve it and control it you will meet the GD&T callouts. I understand that occassionally there are concerns with measuring "too much" but in many cases the same dimensions that are measured for acceptance are also useable for process control; often it's as simple as using each dimension by itself for SPC and collectively for acceptance...

Hard gages cost huge amounts of money and provide zero information, they are used for large runs and as a catch for bad parts, not to make adjustments to a process. They are pretty much junk gages. A hard gage will not tell you how to adjust to make a passable part, only that it fails.

Now that I have thought of it it has been over 10 years since I last saw a hard gage for TP on a shop floor being used. I do find it surprising anyone uses them still. I guess there is a need if you have huge runs and no cmms.

I have seen them in operations where the time it took to get the part to a cmm, get it cued into the workload and get a report was too long or the product pricing structure did not support the overhead of a cmm. When used that way, you can - in a gross sense - set them up to tell when an operator to make an adjustment, and to a lesser degree how far. You make the gage at 75% of the zone the tolerance calls out. Of course, MMC is a critical component of even attempting to use this approach. But, it is far wiser than gages at full tolerance (only good for sorting, and complicated with "fit" issues not found on a cmm). Less helpful than variable data - and certainly not considered for SPC.

The right answer to any question: "it depends"

You make the gage at 75% of the zone the tolerance calls out. Of course, MMC is a critical component of even attempting to use this approach. But, it is far wiser than gages at full tolerance (only good for sorting, and complicated with "fit" issues not found on a cmm). Less helpful than variable data - and certainly not considered for SPC.

The right answer to any question: "it depends"

Gauges should be made at MMC or virtual condition size using 10% of part tolerance as per ASME Y14.43-2003. One would then decide how to apply the 10% gauge tolerance. In the automotive industry, one should apply the tolerance to never accept a nonconforming product.

Gauges should be made at MMC or virtual condition size using 10% of part tolerance as per ASME Y14.43-2003. One would then decide how to apply the 10% gauge tolerance. In the automotive industry, one should apply the tolerance to never accept a nonconforming product.

You always have the option to make a gage a tighter (within) tolerance if you wish to do so as a process control - as in 75% of the accepted zone rather than the full allowed zone. The resulting product will meet a gage made to ASME Y14.43-2003 most readily.

For final inspection purposes, you should have a gage that meets ASME Y14.43-2003 - as final inspection should be considered the equivalent of the customers dock. Having one will calm any customer auditors, and will tempt production to use the more open tolerance gage to circumvent the process control gage. Always good justification for a gage...

Again - we must separate product acceptance from process control. This is where this discussion is going off track. They are not the same thing nor are they mutually exclusive; although effective process control (that includes driving process improvements) will lead to virtually 100% product conformance with minimal acceptance inspection.

we are in the 21st century now...

:mad: even tho I realize many of our community are suppliers to the automotive industry, we do not have to be hostage to only doing what we think they are dictating. After all, they are not stellar examples of a good business an dhopefully we will be around long after their demise or hopefully resurrection... We control our destiny. We are supposed to leaders of improvement; not the guardians of the (long) past norm of inspect and scrap. No one can afford that. And AIAG, no matter how misguided in execution, was well intended to drive us toward process control and improvement and away from merely inspecting and rejecting.

Maybe in the 10 years since I had to deal with hard gages they have gotten better. I have never thought of them as anything but a waste. CMMs used for SPC are very common in large manufactures.

Bev you tell'em. No reason to ever have to sort if you are useing SPC to monitor and control the process. :yes:

My day is over......:biglaugh:

CMMs used for SPC are very common in large manufactures.


OK...that may be true. But let's not suffer from self-reference criterion. Small companies do not always have the means to shuffle over to Wal-Mart and grab a CMM or two. Therefore, they may not be as common in that realm. Does that mean they should walk away from any part with GD&T on the print? I doubt it. :cool:

Dave,

Let’s try again, hopefully simpler this time. SPC never was about verifying conformance to specifications… it was about variation reduction and process improvement! Lately the acronym morphed to include both but it was never intended to do so at the onset. You should know this; you have accrued some years of experience and are you teaching this stuff!

In your 10X hole pattern example… how would you measure the hole’s “in process” parameters to give meaningful feedback to the operator, tool setter, whistle blower, yada, yada, yada,… to signal that something has changed from what was expected…or to reveal opportunities for process improvement.?

First you would look at the process!

If the holes are installed using a rotary table 1x1 or 2x2 or 5x5 it would be intuitive to choose polar coordinates as process control monitoring parameters. If the secondary datum feature |B| was fixtured coaxial with the axis of the rotary table then its position deviation relative to the 10X pattern (as a group) might indicate (half) of how far off the center of the rotary table is to the drilling spindle’s X0,Y0 axis. If the greatest variation was due to indexing then identifying which sets of features belong to the 2x2 or 5x5 would be critical to understanding the index error.

If the holes are installed on a CNC machining center or with a dedicated 10X machining head then it would be intuitive that the individual X, Y, coordinates may be more significant. The 10X pattern’s average position relative to the dedicated head would indicate how far off it was from the fixture that captures |B|, and if |B| was installed in the same operation as the 10X pattern (CNC) the errors would likely reveal the precision of the CNC machine itself.

Notice that we have not talked about B @ MMC or any tertiary datum feature @ MMC this is SPC… namely process control. If one wanted to dumb things down for the operator to an attribute check for process control then the best alternative would be as Bob Doering suggested by making the attribute gage with “less tolerance” than is given in the specification. In doing so the operator can distinguish when the operation is deteriorating and if it remains unadjusted it will likely produce defective product.

Say that we have done all of these fancy things “in process” … our process is in control, and we have set control limits for adjustment to truncate our variation but we cannot reduce the variation, and only 31/32nds with 100% inspection of our product pass the final attribute gauge that you designed. Our capability sucks even if we estimate our capability from our pass/fail performance. The problem is simply that tolerances are too tight to meet capability targets!!! What do we do?

Oh the tolerance is @ MMC :)?

How can that help us?

Old tool makers, maybe older than us, knew that when the parts didn’t fit the gages that they could increase the size of the holes to pass the gauge… but sure enough someone is going to demand the same lofty capability on feature size as well. We’re screwed!

Dave, see a previous discussions to address this concern!
http://elsmar.com/Forums/showthread.php?t=16607

The point is, as Justincredible, Bev, Bobdoering pointed out… attribute gages “in process” reveal little or nothing for the operator to monitor, adjust, or improve the process and as a final inspection weapon when a part fails you have got big problems to meet the capability targets required these days!!!

Your buddy.

Paul

OK...that may be true. But let's not suffer from self-reference criterion. Small companies do not always have the means to shuffle over to Wal-Mart and grab a CMM or two. Therefore, they may not be as common in that realm. Does that mean they should walk away from any part with GD&T on the print? I doubt it. :cool:

A QC rep should be involved and have the needed knowledge to ensure the print and all supporting documentation can be done per customer agreements.

It is not very good if a salesman gets a job that can not be checked, or gets a job that reqs SPC and there is not the capibility to meet the contract.

GD&T can be checked with more than a CMM, and SPC can be done on GD&T callouts not on a CMM. "The rock never lies", the majority of the time I will recheck anything out on the cmm with hand gages. As a matter of fact I have a part running right now next to me that I did not accept the paraelleism reading on the CMM so I rechecked with a mic and accepted the parts. I will do the same if the any of the features read out on the CMM. There is nothing that a CMM checks that can not be done with hand gages.

Paul:

I did not say that SPC could not be used on positional tolerances. I said that it was not practical since it does not cover all the simultaneous requirements.

If it is helpful to plot centres on the 10 holes in RFS relative to datums in RFS, go for it. It is costly though. Imagine, once an hour taking a part to the CMM room, tearing down whatever is on the table to check centres. We would have a sub-group size of 10 but would require two (2) charts - one for the X direction and one for the Y direction. There would not be a correlation to the specification though.

I remember when we started utilizing SPC on the shop floor in the early 80s, we had to come up with methods so the Operator would take no more than 3-4 minutes to measure and plot the chart since they are paid to make product. The "critical" characteristics were checked on a variable checking fixture with a set block. It was paramount to keep the inspection time low. Should the Operator take the part to the CMM room? Maybe we should have the Tech take it? If we had the Operator wait to have the holes measured, we would lose our time cycle and Operator efficiency.

Statistically plotting 10 centres can absolutely be done and I would consider the characteristic a process characteristic rather than a product characteristic.

Is it possible the we could have the centres in statistical control but maybe 1 hole is quite a bit off 90 degrees and the part does not conform to the requirements? Is it possible that a Customer rejected a shipment of product since some product would not assemble and we did not confirm that it met all the simultaneous requirements included in the positional tolerances?

If a company desires to statistically plot centres, I would suggest it also confirm that the positional tolerances with an attribute checking fixture produced as per ASME Y14.43 - 2003. This might be the best of both worlds.

Respectfully,

A QC rep should be involved and have the needed knowledge to ensure the print and all supporting documentation can be done per customer agreements.

I agree. Throw a CMM in the quote-no problem.

It is not very good if a salesman gets a job that can not be checked, or gets a job that reqs SPC and there is not the capability to meet the contract.

I agree. If there is a contractual agreement to do SPC, then it should be done (even if it is attribute). And, yes, the job should be able to be checked as a part of the contract review. Checking may be with a hard gage.

GD&T can be checked with more than a CMM, and SPC can be done on GD&T callouts not on a CMM. "The rock never lies", the majority of the time I will recheck anything out on the cmm with hand gages. As a matter of fact I have a part running right now next to me that I did not accept the parallelism reading on the CMM so I rechecked with a mic and accepted the parts. I will do the same if the any of the features read out on the CMM. There is nothing that a CMM checks that can not be done with hand gages.

I agree. If the operator you can lay out the dimension on a rock next to the machine within the cycle time - go for it. Heck, I always said if you want quality to determine the measurement requirements, check 100% of the dimensions 100% of the time. Plot them all, too. If you pass that you can get sleep at night that no parts are coming back. But, there seems to be some resistance to that approach. I never said that a hard gage was the only way or preferred way, but there are still many cases where it is sufficient (cost effective?) and can assure that bad parts are not made (tighter tolerance gage). Those cases may not be the elegant example of a bolt pattern in space, but more like a true position of one diameter to another to ensure fit.


By the way, I have been many places where the CMM is not trusted by the operators because it does not duplicate layout results. Makes for great conversations on the shop floor. Great example!!