Six Sigma on Process Control

Quote from Wikipedia: "Six Sigma's implicit goal is to improve all processes, but not to the 3.4 DPMO level necessarily. Organizations need to determine an appropriate sigma level for each of their most important processes and strive to achieve these. As a result of this goal, it is incumbent on management of the organization to prioritize areas of improvement."

What that tells us is that some processes may require a higher sigma and others may be just fine with a lower sigma deviation. The trick is knowing what is and what is not important to the overall picture of your quality needs and drive for that.

The biggest problem with 6 sigma is who invented it, who implements it and who enforces it.

It was invented by people who don't make anything and implemented by MBA's who didn't understand how things really get made.

And if it is enforced by people who want to be enforcers and not facilitators, it can be costly.

Case in point. Injection molded plastic parts. Each part may have 20-100 attributes and you may have 10,000 parts.

Applying 6 sigma to each attribute will not facilitate acceptable parts, it will increase DPMO.

Applying 6 sigma to the processes used to make the parts can be beneficial.

As Anonymous Hero said, if all players understand the real goals it will help. If not, it may not help.

The biggest thing I got out of the 6 sigma class they pushed back 12-15 or so years ago, was to make your design in sensitive to individual component tolerances. So if your working with some electronic circuit, you do a sensitivity analysis on the transfer function to each part, and select component values (or total design change) to reduce the sensitivity of the design to component variances.

It was taught by some traveling instructor from the Motorola before they sold off their silicon foundry and design group. One of the examples was an notch filter, and they mapped a 3D graph of 2 components with the greatest sensitivity. They showed picking the minima of this graph surface.

Also expect the 6 sigma to apply to each manufacturing process you put the product through, point being even if each process was 6 sigma, you may have so many processes applied that you still get zero output.

At least that's what I got out of the class.

3 years ago, just before I retired, they were pushing 6 sigma again. And they had folks in production that got titles of Queen/King of 6 sigma. The problem was they had no data. Most test outputs were boolean, PASS/FAIL, with no analog measurement of actual output value (e.g. a discrete output, for open/+28, PASS if greater then 16V, but where did it saturate at, and what was the load it drove). It was always funny that test equipment was relegated to the new kids, fresh out of school. WTF did they know about product test? The quality was a function of how sharp the kid was. History showed many service bulletins to the test equipment that failed to test the product in the same way it was wired in the aircraft (no attempt at simulating the worst source or load impedances). Worse is the same test equipment used in the factory, was in the service centers. so guess what, No Fault Found (NFF). I guess that was OK as one SB fixed them everywhere.

The point is applicable to mechanical tests using a go-nogo gauge, You need to make precision measurements of fit if your going to fix a machining problem. i.e. you need detailed measurement data for any 6 sigma process to be analyzed.

I lived through the Hell of 6 Sigma indoctrination at Motorola in 1986/7/8.

As I said earlier, it's fine, if done correctly and for the right reasons.

As someone who is older been in the profession longer than Six Sigma has existed, I entirely embrace this quote:

From what I've seen of (Six Sigma), it's a basic version of quality improvement. There is nothing new there. It includes what we used to call facilitators. They've adopted more flamboyant terms, like belts with different colors. I think that concept has merit to set apart, to create specialists who can be very helpful. Again, that's not a new idea. The American Society for Quality long ago established certificates, such as for reliability engineers.
― Joseph M Juran

SO my question is if my real time data is within that range(six sigma range) only, will it be considered as good data?? Or this wont be applicable for process real time loops.. if we can apply then wat is the range for it??

Real time data is only going to be partially important without a historical trend to it.

However, once the upper and lower limits are statistically established via Six Sigma tools, you can use real time data to determine if the part is good or bad.

The thing is, without historical data, when a part exceeds the limits you do not know if the problem is simply an outlying flyer or part of a trend.

Historical data tracking shows you if the process is steering away from the ideal and you can avert a line shutdown before you start producing bad parts. That's the key takeaway here. You can't do that if you are simply applying a go/nogo test to each part.

Your real data is what it is. You define what it should be and go through the process of evaluation, verification and process steering based on the facts.

What real time data are we talking about?

Data is neither good nor bad.

The control of processes is always asking the question, "Is there a statistically significant difference in my process between the last time I checked it and now?" If the answer is "No", then the process is stable and its outputs should be conforming (good). If the answer is "Yes", then the process is not stable and is likely producing outputs that are nonconforming (bad). The power in the statistical testing is that your sample data can show this process shift before your sampling plan shows you a nonconforming part.

A major useful feature of SPC is that it allows you to gather your data for this process run and, without having to perform a t-test or ANOVA, look at your chart and know you are still stable. An X-bar R chart is one of the simplest things to set up and use, and only requires the operator to be able to add, subtract, and divide. Further, the information from the data is available before the next process run starts, so you don't keep making nonconforming material. Even better, it can allow you to correct a drifting process before the process produces any nonconforming material.

All quality systems are to be develop to your organization. This can be modified for you to handle/control your quality issues. For which you develop your standards.

You 6σ program is to compliment your organization, to be develop around your company needs and requirements.

Don't confuse it with ISO.

It depends on what the process is and how finite the allowable PV control error is.

The PV control tolerance is critical for manufacturing medicines, chemicals, and other critical processes while maintaining an exact level in a storage tank for one of the chemicals/fluids is not.

GA. I tried saying it differently below.

Thanks for the GA.

The reason that I suggested further reading is that the OP seems to be using "batch" concepts of sigma when it would seem that there is really a process happening. He should learn about "run charts" as opposed to just the basic stats.

It is possible to have data that provides mean, SD etc that is actually trending over time (unstable).

So much to learn!!

"Wat i mean to ask is like if all the real time data values lies within that range only then that loop is good??"

The first thing you need to do is understand how that process behaves. A process will have its own pattern which, by itself, is neither good or bad. Let the process run without adjustments. Gather your data. Rule of thumb, at least 30 data points. Calculate mean and std dev. Your process is likely to run at mean +/- 3 std dev. If these values are entirely inside your upper and lower spec limits, you may have a process which is usable. If any part of that +/- 3 std dev spread is outside of a spec limit, you definitely have an unusable process.

After you know this, you can use things like 6 Sigma to make it better.

Also, I am not sure exactly what you mean when you say "the six sigma concept." Let me try to clarify what we mean by 6 sigma.

6 sigma, with a lower case "s", +/- 3 standard deviations, is the foundation of Statistical Process Control, or SPC. This is the work of Shewhart and, while it began in 1924, is rarely used as Shewhart intended.

6 Sigma, with a capital "S", is a trademarked discipline. Confusingly, 6 Sigma bases its philosophy on +/- 4.5 standard deviations to allow the process to drift before producing an unacceptable amount of nonconforming material. The definition of the 6 Sigma concept has changed multiple times over its history, so people are always arguing over what it really means. Like SPC, 6 Sigma is rarely used as it was originally intended.

SPC is a useful tool because it helps you decide if a nonconformance is a result of process noise (random) or process drift (non-random). If process drift, solutions tend to be quick, easy, and inexpensive. If process noise, 2 times out of 1,000 for +/- 3 standard deviations, you ignore it unless, for some reason, this is unacceptable. Fixing process noise tends to be slow, difficult, and expensive. These fixes must be fully supported by management with dedicated money and people and time.

6 Sigma, with a capital "S", insists on +/-4.5 standard deviations. This means nonconformances due to process noise can be expected about 3 times out of 1,000,000. This is justifiable on certain products and processes, but, for most products, is not reasonable to implement on 100% of your processes.

100% implementation of either SPC or Six Sigma is not always necessary. You must use your time, money, and people wisely, so implement controls where you most need them first. This is the Pareto Principle, one of the basic Quality Tools. Also, you might improve a process so that SPC is no longer required.