We are a small to medium size metal stamping company. As of today, we do no SPC data collection. Our QS auditor has been hinting that we need to collect data to prove that we meet or exceed the capability at PPAP. Currently, we use attribute gages, go/nogo feeler checks to approve, reject parts. We do cap. studies for ppap's. In taking a refresher class, I was talking with the instructor, and he said that we could use the X-bar, moving range charts on first and last pieces. In the opinions of the board here, would this give any beneficial information to our company?

Welcome to the Cove, Larry!http://elsmar.com/Forums/images/smilies/bigwave-d2.gif

Xbar charts are wonderful for many production runs at helping determine whether your process is in control.

SPC is a mathematic manipulation of the data to generate a number to report how "tight" that control is.

What's a typical production run of a part in number of pieces, Larry? If it's too small, SPC may not have any probative value to justify the time and expense. Have your customers provided one or more critical characteristics which should be reported for SPC purposes? If not, can you determine one or more characteristics of the finished piece part which most accurately reflect whether your entire process is in control? If yes, those are the characteristics to chart for SPC purposes.

production runs vary in length, best guess would be 5000 pieces. Concerning the critical characteristics, they supply none. We picked what we determined would a dimension that would be critical to the function of the bracket at the customers facility

production runs vary in length, best guess would be 5000 pieces. Concerning the critical characteristics, they supply none. We picked what we determined would a dimension that would be critical to the function of the bracket at the customers facilityConfirm the criticality with the customer to assure the SPC figure will be meaningful and useful to him.

Remember this: Some customers may demand SPC charts, but the real value of SPC is to the producer to help him understand how tight his control is.

My bad, the instructor told me that the X & moving R chart was be a good tool for use to use on only the first and last piece of a production run of our stampings. Does this seem like a good choice of statistical tools for us to use?

My bad, the instructor told me that the X & moving R chart was be a good tool for use to use on only the first and last piece of a production run of our stampings. Does this seem like a good choice of statistical tools for us to use?I was talking about an Xbar chart. Perhaps the world has passed me by, but we used Xbar charts from beginning to end of a production run. It was not to generate an SPC number, but to determine simply if our process was in control.

Here are just some of the different charts which may have an impact on determining whether and how much your process is in control:

X Moving Range
XBarR
XBarS
NP
C
U
Z scale

Sometimes, different writers call some of them by slightly different names. The synonyms for an Xbar chart (for example) include:

XBAR CONTROL CHART,
MEAN CONTROL CHART,
MEAN CHART,
X CHART,
AVERAGE CONTROL CHART, and
AVERAGE CHART

Moderator edit: [10-1-2009] link to a white paper on SPC was broken and thus reference removed]

My bad, the instructor told me that the X & moving R chart was be a good tool for use to use on only the first and last piece of a production run of our stampings. Does this seem like a good choice of statistical tools for us to use?

I don't know the basis for the "first and last" statement. That seems very odd. If you have data anyway for all pieces, all should be included. Now, if you go to xbar-R and do rational subgroupings (most commonly sets of 4 or 5 sequential parts) then the chart doesn't get quite so huge.

It is an interesting discussion of what happens when you are "forced" to do SPC. I would not have my current position nor my interesting Hanford history had not the government insisted to Westinghouse that they have me start making control charts. Many times the immediate need is external. But control charts without the "statistical thinking" and better, the "Systems thinking" to go with them will just make them pretty wallpaper. But, unless you make the initial leap to use SPC, you will never achieve statistical thinking nor systems thinking.

It is an interesting discussion of what happens when you are "forced" to do SPC. .....control charts without the "statistical thinking" and better, the "Systems thinking" to go with them will just make them pretty wallpaper. But, unless you make the initial leap to use SPC, you will never achieve statistical thinking nor systems thinking.

Steve rightly wants you to plunge in; even if the initial leap is through the "first and last" piece inspection. But choose them only when the production process has stabilised and not literally. ( for the conforming pieces within specification limits). Infact for destructive tests performed some organisations are seen resorting to this practice, with a moving range of 2 samples.

SPC is an ocean. The water is not yet too cold to swim.
thareja

Larry, You said earlier that you currently do attribute checks on your stampings. From that statement I gather that many of the specifications that you are dealing with have MMC (maximum material condition) modifiers applied to them.

I would just caution you that after you begin your process monitoring and control procedures and reveal the predictable common cause variation someone will likely require you to estimate the capability of your process to the customer specifications. If you try to do that with your continuous data you must recognize that the equations for Ppk and Cpk accept only a constant value of tolerance (USL) and if there are MMC modifiers applied to the geometric tolerance the limit is variable (with respect to feature size).

If you try to figure Cpk with the continuous data the variable portion of tolerance will be ignored and your process capability will appear worse than it actually is. Your attribute gages are built to verify the boundary (virtual condition) of that variable tolerance so that is the only right way to predict the capability of your process to the specification. Just tell the auditor that you can't predict Cpk with variable tolerances!

I have a similar situation i am supposed to deal with. I work at a reel to reel metal stamping company meaning we can sample only either at the beginning or at the end. We are not supposed to cut in between a reel. Its very hard to install any measuring gauges, even if possible it will not be in near future. How can this process be monitored for control?

I have a similar situation i am supposed to deal with. I work at a reel to reel metal stamping company meaning we can sample only either at the beginning or at the end. We are not supposed to cut in between a reel. Its very hard to install any measuring gauges, even if possible it will not be in near future. How can this process be monitored for control?
The key to effective SPC is rational subgrouping and the frequency between subgroups.

To do both correctly, you need to analyze your process for the dominant sources of variation. (Note: AIAG has a good section on X-Dominant processes in their APQP manual). Depending on the type of stamping that you are doing, your process is typically either a tooling-dominant or setup-dominant process. This means that the major sources of variation are either from the tooling (e.g., wear, damage, etc.) or from the setup (e.g., die positioning, spacing, strokes per minute, etc.).

In either case, changes occur slowly (i.e., die wear or setup) unless something catastrophic occurs (damage) I see two possibilities for SPC here:


Xbar/R (or S) chart using the 1st/last pieces for a subgroup size of two. Within subgroup variation is within run variation. Between subgroup variation is run-to-run variation. You will primarily detect setup to setup variation and tool wear.
X MR chart using 1st piece from each run for a subgroup size of one. Your limits will be based on setup-to-setup variation, so this will primarily detect tool wear unless a particular setup is highly variable. (Note: I do not recommend using the X MR chart with a sequence of 1st / last / 1st / last / etc. This is alternating different sources of variation and could result in unusual patterns.

Thank you for replying. This is my first SPC project at my new job. I will do some research on the lead you have given. If i have some questions, i will come back and post here.
Thanks a lot!!

Azzuri, are you able to collect parts periodically? And about how long are your runs?

cjkurtz, i am not able to take any kind of samples as we can not cut in between the reel. However, we can take out pins, which are punched into the vandelayrs and and inspect them in-process.
The cycle time depends on the design and varies around 2 hrs per reel.
Are you into similar process?

cjkurtz, i am not able to take any kind of samples as we can not cut in between the reel. However, we can take out pins, which are punched into the vandelayrs and and inspect them in-process.

Just for clarification, the word in bold (my doing) above should be bandoliers (http://images.pennnet.com/articles/cs/thm/th_0603cs_feature02_04.gif) (or bandoleers).