I'm sure I'm not the only person that is trying to run an SPC program while having laid people off... and of course, in my experience, QC is always hit hardest because it's "no value added" to the bean counters.

I running into serious concern with maintaining SPC because it's still a baby here and QC does all the inspection and data entry.

But with layoffs I just can't keep up - at least at the sample frequency we want (1 inspection per hour for most parts). Even with half the machines down one inspector can't do SPC + help operators + do final inspection on non SPC parts + do an oocasional first article.

Here's the choices I see:
1) reduce inspection frequency for all but the most difficult parts
2) reduce characteristics measured on the parts and eliminate the more time consuming measurements (characteristics done on a optical comparator or with a surface tests)
3) Get the operators to start doing their own SPC data entry

I'm going to propose 3. We're slow so now is the time to do programs such as this. I'll let you know how it goes.

what are your experiences?

3) Get the operators to start doing their own SPC data entry :applause:

Cha Ching ! :agree1: In my opinion, they should always do their own entries, as they control the process behavior. The Inspectors should Audit for Compliance, but not be the Data Entry Clerks.

For the other inspections, reduce them based upon data illustrating the lowest risk products or families. Just be sure that if your bound to Control Plans based upon Customer Specifics that you don't get into trouble from an Audit Standpoint. Sadly, the Registration Bodies and Customer's don't understand economics, so don't get busted. :2cents: Good Luck----

I would like to offer up some suggestions, but I need some idea of the processes, types of characteristics, the tolerances, etc. The choices you list are really not the best criteria for getting out of your dilemma. That does not mean they may not end up being employed - but if there is some statistical and/or risk basis for reduction of sampling, then you want to use that logic first.

Option 3 is commonly employed, even without the type of situation you have. It should be their responsibility. They own the process and its results, and should be the experts of the process' condition. After all, they have their hands on the process - not you (presumably).

There is no right or wrong solution. What you need to do is determine the risks involved with each of your proposals. If you reduce the frequency of checks, what is the risk of producing a lot of scrap or rework? If the machine operators perform the checks, what is the risk of an inaccurate measurement and non-conforming parts being shipped?

It could be that the best solution is a combination of your proposals.

I would like to offer up some suggestions, but I need some idea of the processes, types of characteristics, the tolerances, etc. The choices you list are really not the best criteria for getting out of your dilemma. That does not mean they may not end up being employed - but if there is some statistical and/or risk basis for reduction of sampling, then you want to use that logic first.

Option 3 is commonly employed, even without the type of situation you have. It should be their responsibility. They own the process and its results, and should be the experts of the process' condition. After all, they have their hands on the process - not you (presumably).


We're a machining operation, swiss type CNC lathes for the most part.
Characteristics are mostly linear dimensions measured with hand tools and optical comparators. Other characteristics are surface finish measured with a pocket surf; threads checked with plug and ring gauges; and visual attributes like burrs, sharp corners, etc.
Tolerances on the linear dims can be from +- .001" to +-.010". If it's over .010" it generally won't be a CTQ characteristic or a predictor of other characteristics.
We've tried to identify the CTQ characteristics by meeting with programmers to see what is made in the same pass and selecting one as a predictor for the other cuts in the same pass.

Problem with getting operators to do SPC is something I've been fighting for over three years now - they are viewed by their managers as transient labor who are not worth putting the training into becuase they'll skip out for for $.50 more.
That's harsh, but it's the truth.

I am not sure if you have had a chance to read the blog Statistical process control for precision machining (http://elsmar.com/Forums/blog.php?b=79), but if you were using correct SPC for machining, you would be able to determine very accurately the SPC frequency- whether it be every five minutes or once per shift. If you are using X-bar R charts for lengths and diameters, then you are really likely wasting your time. As far as which dimensions you should be charting, it should the the tightest dimension for each finishing tool. Your chart should be telling you when to make an offset, and your R chart should be telling you when it is time to change your tool. It won't be, however, if you are using X bar-R charts - and the same pass notion will not be effective with that chart. It is just the worst chart for precision machining.

Doing correct SPC will get more operator buy-in because it makes sense to them, tells them when to make an offset, and will likely have them adjusting much less, if your tool wear is decently good. I recommend using the X hi/lo-R chart for precision machining.

As far as your attribute checks and surface finishes, etc., your process (speeds, feeds, tools, raw material) will have to dictate how frequently you check.

I'm sure I'm not the only person that is trying to run an SPC program while having laid people off... and of course, in my experience, QC is always hit hardest because it's "no value added" to the bean counters.

I running into serious concern with maintaining SPC because it's still a baby here and QC does all the inspection and data entry.

But with layoffs I just can't keep up - at least at the sample frequency we want (1 inspection per hour for most parts). Even with half the machines down one inspector can't do SPC + help operators + do final inspection on non SPC parts + do an oocasional first article.

Here's the choices I see:
1) reduce inspection frequency for all but the most difficult parts
2) reduce characteristics measured on the parts and eliminate the more time consuming measurements (characteristics done on a optical comparator or with a surface tests)
3) Get the operators to start doing their own SPC data entry

I'm going to propose 3. We're slow so now is the time to do programs such as this. I'll let you know how it goes.

what are your experiences?

In the end, always had better experience with Pre Control charting by the operator, works really well and it is very visual so the operator and machinists usually handle it well.

I am careful to try and only use SPC where it really fits and there is dollars and sense benefits!

Good luck!

In the end, always had better experience with Pre Control charting by the operator, works really well and it is very visual so the operator and machinists usually handle it well.

I am careful to try and only use SPC where it really fits and there is dollars and sense benefits!

Good luck!

One of the problems with Pre-control is it is based on the normal distribution. The normal distribution in precision machining usually indicates a process out of control, and typically over adjusted and tainted with measurement (not gage) error. Precontrol tends to be the cause of the problem, and not the fix for the problem. It is simply the wrong chart for the process. It won't tell you it is the wrong chart, though, so you think it is working.

Using the wrong charting methodology has caused most of the frustrations that have made people think in terms of dollars and cents, and not process control, statistically supported sampling frequencies and meaningful feedback from the charting.

The concept of pre-control charting is an interesting concept for precision machining. One reason that it will appear to work in precision machining is that the technique is almost half correct. Some use a of set control limits based on a percentage of the specifications (typically 50%) has been used as a starting point for collecting data that may ultimately be used to generate an X-bar-R chart or I-MR chart. Some folks may not make that leap, and stay at the precontrol limits, because the seem more reasonable than the limits calculated with the other two charts. The reason why that is true for precision machining is that the calculations for the correct continuous uniform distribution are calculated based on a percentage of the specifications. So, it makes some sense, although if correctly controlled one could likely expand the control limits to 75% of the tolerance. Others develop control limits based on standard deviations from a capability study, and that will be of no value whatsoever.

The half that is incorrect is that the data plotted on the chart are typically averages of 3 or 5 parts - maybe even individual measurements. The averages are an abject waste of time, and the individuals are only good if the roundness (for round characteristics) or parallelism (for lengths) is insignificant compared to the specification (as in about less than 10%). The most correct would be the X-hi/lo-R charting methodology (http://elsmar.com/Forums/blog.php?b=79), which includes setting an OD to the lower control limit, allowing the tool to wear to the upper control limit, than readjusting back down to the lower control limit (opposite for ID), generating the sawtooth curve and the correct distribution. Adjusting to the mean will give a normal distribution based on the operator adjustment - not the actual process (which is tool wear) - exhibiting classic overcontrol and process tampering.