Precontrol Charts for unilateral tolerances on 'new' processes

[SteveL]

We are introducing a series of Precontrol charts on 'new' processes in a precision machine shop so we can 'manage' the process until such time we have enough data to apply Control Charts.

Bilateral tolerances appear straightforward (Green Mid 50%, etc) but I'm unsure about what to apply for unilateral tolerances. Can you actually apply precontrol charts to single-sided tolerances as they are normally 'skewed' to the right in terms of shape

Many thanks in advance and apologies in advance if this has been answered elsewhere.

 

[bobdoering]

Re: Precontrol Charts for unilateral tolerances

We are introducing a series of Precontrol charts on 'new' processes in a precision machine shop so we can 'manage' the process until such time we have enough data to apply Control Charts.

Bilateral tolerances appear straightforward (Green Mid 50%, etc) but I'm unsure about what to apply for unilateral tolerances. Can you actually apply precontrol charts to single-sided tolerances as they are normally 'skewed' to the right in terms of shape

Many thanks in advance and apologies in advance if this has been answered elsewhere.

Are these dimensional unilateral or form (roundness, flatness, etc.)?

Also, for correct SPC in precision machining, you may want to consider the X hi/lo-R chart:
http://elsmar.com/Forums/showpost.php?p=188214&postcount=3

Precontrol is a mean-oriented control chart, and for precision machining (diameters, lengths) they are non-normal uniform distribution. They are not a good chart for precision machining - they encourage overcontrol. The mean has no meaning to those types of dimensions if charted correctly.

 

[David DeLong]

Re: Precontrol Charts for unilateral tolerances

We are introducing a series of Precontrol charts on 'new' processes in a precision machine shop so we can 'manage' the process until such time we have enough data to apply Control Charts.

Bilateral tolerances appear straightforward (Green Mid 50%, etc) but I'm unsure about what to apply for unilateral tolerances. Can you actually apply precontrol charts to single-sided tolerances as they are normally 'skewed' to the right in terms of shape

Many thanks in advance and apologies in advance if this has been answered elsewhere.

I am going to assume that you have taken short term capability studies and if you have, maybe a Modified Control Limit chart would be best. This chart reflects the averages making sure that you have a 99.73% probably of making parts within specification. It really doesn't get into runs, trends etc. but is a good starter chart. After, say, 25 entries, one has enough data to begin running X bar and R charts assuming normality using the information from this Modifiied Control Limit chart.

The Range chart would also be used but without control limits.

I picked this up so many years ago attending the 5 day Ford SPC (you better attend) seminar in Detroit and is shown in Statistical Quality Control by Grant & Levenworth. I a Quality Manager of a Ford supplier at that time.

Basically you need to calculate the estimate of the standard deviation from your short term study. The formula for the UCL modified is USL - (V X SD') where V is 1.66 for a 5 piece sample size and varies depending upon the sample size. If averages are below the UCL modied, then the product is within specification. If an average is outside the UCL modied, shut down and sort back to the last check. Some people used to call these "Shut Down Limits".

Hope this helps.

 

[bobdoering]

Re: Precontrol Charts for unilateral tolerances

I am going to assume that you have taken short term capability studies and if you have, maybe a Modified Control Limit chart would be best. This chart reflects the averages making sure that you have a 99.73% probably of making parts within specification.

Actually, with precision machining, you are better off setting you control limits to 75% of your tolerance. That is if you are truly doing precision machining, which means the main variation of your process is tool wear. Again, other control chart methodologies are not effective when properly applied to precision machining because of its uniform distribution. X-bar and R charts are the absolute worst - but this has been discussed elsewhere in the forums (see link above) and the following thread: http://elsmar.com/Forums/showthread.php?t=20935&highlight=bobdoering

 

[Mike S.]

Re: Precontrol Charts for unilateral tolerances

We are introducing a series of Precontrol charts on 'new' processes in a precision machine shop so we can 'manage' the process until such time we have enough data to apply Control Charts.

Bilateral tolerances appear straightforward (Green Mid 50%, etc) but I'm unsure about what to apply for unilateral tolerances.

For unilateral tolerance...

red zone = OOT
yellow zone 1/4 tolerance
green zone = 3/4 tolerance

 

[UK Steve]

Re: Precontrol Charts for unilateral tolerances

Are these dimensional unilateral or form (roundness, flatness, etc.)?

Also, for correct SPC in precision machining, you may want to consider the X hi/lo-R chart:


Precontrol is a mean-oriented control chart, and for precision machining (diameters, lengths) they are non-normal uniform distribution. They are not a good chart for precision machining - they encourage overcontrol. The mean has no meaning to those types of dimensions if charted correctly.

Thanks for replying Bob

The dimensions in question will be a mixture of flatness, concentricity, perpendicularity and parallelism - which tend to machine-related geometrics.

Interesting you mention that in precision machining features such as Diameters and Lengths appear non-normal. The only time I've found that is in long term analysis when tools have exhibited wear over time and been reset -hence the distibution appears bi-modal. Otherwise they not only appear normal but an AD tests confirm their normalty too. As we have over 2,000 processes (yes I know paralysis through analysis) being charted I can gladly show examples.

The point I'm still unclear about though is do Precontrol charts exist for skewed distributions and if so how are they categorised?

 

[UK Steve]

Re: Precontrol Charts for unilateral tolerances

For unilateral tolerance...

red zone = OOT
yellow zone 1/4 tolerance
green zone = 3/4 tolerance

Sorry Mike just missed your post!

It's as simple as that eh!

I've searched high and low for docmentation about unilaterally toleranced Precontrol charts and found nothing. Have you any links I can use to demonstrate to our Engineering Dept that they do exist and are valid?

 

[Bev D]

Re: Precontrol Charts for unilateral tolerances

The point I'm still unclear about though is do Precontrol charts exist for skewed distributions and if so how are they categorised?

As you and Mike S posted at about hte same time - Mike S has the formulas for you. The will work for the types of distributions you mentioned, IF you meet the other criteria: the ability to truly control the process to your target zone is proven. The difference between tampering and control is validated knowledge of what factors casue the variation and the ability to reliably change them to achieve the desired targeted result. PreControl is really a manual version of "Engineering Process Control" or "Automated Process Control" where algorithms and automatic sensors monitor processes and adjust input factors without human intervention - in fact most of us have these in our homes, they're called temperature controls...

Some processes are perfectly suited for PreControl and others are not. For example Bob D's application for a uniform process (and yes you must first confirm that uniform process exists in the relatively short run and it's due to a systemic cause - tool wear) is a hybrid approach that works: it's essentially more aligned with PreControl than it is SPC.

There are MANY tools at our disposal. No one tool - tho may be your personal favorite - should ever be blindly or universally applied. It is our job to understand our systems and our needs in accordance with the laws of physics and statistics and select the correct tool to meet those needs.

 

[Bev D]

Re: Precontrol Charts for unilateral tolerances

Sorry Mike just missed your post!

It's as simple as that eh!

I've searched high and low for docmentation about unilaterally toleranced Precontrol charts and found nothing. Have you any links I can use to demonstrate to our Engineering Dept that they do exist and are valid?

Shainin, Dorian, “Better Than Good Old X, R Charts asked by Vendees”, ASQC Quality Congress Transactions - Chicago, 1984

it is probably also described in Juran's Quality Manual but my copy is not accessable right now. (water pipe burst above my new office books in large heavy totes now...)

 

[Atul Khandekar]

Re: Precontrol Charts for unilateral tolerances

Check: http://symphonytech.com/articles/precontrol.htm