Process Capability

You might want to start with an equipment list of what you have available to work with. Sizes, materials of construction, ratings, etc.

As for figuring out what to do with that, get in contact with a contract manufactuing company and see if what you have matches up with something they want to manufacture.

There is a lot of information regarding process capability. What are the specifics?

i am into piston manufacturing.therefore tell me

Try searching the web for things like 'standard deviation','histograms','sample size' etc.

What you are trying to establish is the fact that the process, (which starts with the quality of the raw materials, the machine and tooling, the actual machining operation, and any other operation encompassed by the process,) is a repeatable, reliable, and effective method of production.

This is a calculable figure. For each part of the process, a sample is taken and measurements are taken! These measurements could be the diameter of a critical journal, thickness of plating, surface finish, ovality, what ever you need to analyze.

With the tolerances specified on the spec, you can calculate using standard deviation, histograms, and other formula, the range with which your process can operate. If your tolerance for a journal is 100mm dia ±0.2mm and your machine was knackered, when you measured the sample (with out adjusting the machine ) you would find a wide spread over the whole tolerance band, maybe with a few strays out side the limits! Your sample is telling you that the process as it is now, is not reliable, repeatable and not the way to go! If you then had the machine serviced and restored to top notch, you would repeat the process, taking your sample and measuring the said journal! This time, you find all the measurements fall within the tolerance band, great!! but that's not the end of it! To be a capable process, it needs to not only be in the tolerance band, but occupy only 30% of the band! ( off the top of my head, think it 0.303 of a standard div, but I may just be blowing smoke!!) If your process can hold this 30% tolerance, then it is normally deemed a capable process! But again, that is not the end of it! You will need to take more samples through out the run! this is good because it can show trends in the process such as preventative measures! Eg, the turning tool lasts 500 parts then needs changing instead of 525 parts made and it's time to check all the parts back to when the tip broke! It can be a very helpful tool if embraced in the right manner!

Check out this for a few ideas! ( I haven't read it all, but on first impressions, might be of use!)

Good luck

Funny you mentioned bat spew into gold. If you would send me $29.95 American, I will send you my special, top secret, classified, spew to gold conversion diagram and worksheet.

It's guaranteed.

Capability studies use indices (index numbers) to represent this performance. The general short term formula is: Cp=(USL-LSL)/6δ; there are others but this will cost you $95.00/hr if I continue...

Are you asking about process capability meaning the ratio of variation inherent in the process in comparison to the blueprint specifications of the part being produced?

If so, Mr. Truman Brain and Ronald got you started, and this link, http://www.qualityamerica.com/knowledgecente/articles/cqeIVH4_1.htm

and possibly some of the other documents on that site will describe what you need to do with your process equipment and measurement system before you will be ready to collect the data needed in order to use the formula Ronald provided. Which, by the way, is only valid if you have targeted your process to produce parts in the middle of a symmetrical blueprint spec, like + / - 0.0N mm. The case where some specs are +0.0N, -0.00 or vice-versa ("must not be under, but over a bit is OK" or "must not be over, but under a bit is OK") requires a different formula.

There is also a similar index, called Z in the Six Sigma methodology, but if you aren't in a Six Sigma organization, you don't need the complexity of that metric, and if you were in a Six Sigma organization, you wouldn't have been asking CR4 about "process capability".

I too am a "quality" consultant, at least some of the time. But I don't mind pointing beginners towards helpful information when I can. If you need a lot more help than these hints and a textbook / handbook on SPC, try the forum on www.asqc.com

And if it turns out you're asking about some of the other sorts of process capability, the formula varies with the process, so you'll have to tell us quite a lot more about what you're doing before we can help you.

Anna

Process Capability = Cp = (T/6σ)

where, T = Total tolerance band

σ = Standard deviation

σ = √((∑(X_n-X)²)/n)

where, Xn = measurand value

X = Arithmetic mean

n = sample size

Process capability should always be greater than 1

If 6σ = T, then the process is just capable

6σ > T, the process is not capable

6σ < T, the process is hoghly capable

But, Cp should not be the sole decision maker, there is another factor called Cpk (Capability index or Centering index)

Cpk = Min of [((Upper specification limit - Mean)/3σ), ((Lower specification limit - Mean)/3σ)

Cpk should be greater than 1.2

This indicates centering of the process, which is as important as Process capability.

Thank You Prashant,

Have dinner on me!

"Cpk should be greater than 1.2"

Shouldn't that be 1.33?

The one company I worked for that used it very efficiently, had the following rule of thumb (if designed using 6sigma):

Cpk=2.0 Target

Cpk<1.66 Begin investigating, but let process run

Cpk<1.33 Shut process down & fix problem

Mind you now, the company was making automotive crash sensors so some might think that is a little too strict, but if you're target goal is zero defects & a Cpk of 1.00 will guarantee you rejects, then this rule makes sense.

To the original question;

These formulas that were posted can be very effective, as long as you are choosing collecting the proper information to use.

2 differing examples;

Company #1: Electro-mechanical crash sensor that is entirely resistive welded has 2 samples taken at the start of every shift and depending on life of the electrode, sampled periodically (i.e. every 200 or 400). Samples originally were measured by weld nugget size, which became too subjective, and was changed to pull or shear force test depending on weld.

Company #2: Collected final test data at the low volume engineering stage & used Cpk values to 'sell' capability to production. I used the term sell because that is all it was useful for considering 100% of the product was tested & gross failures that were explainable were removed from data, hence, no practical use for statistical analysis other than selling someone a 'bag of goods'

For another point of view, one of the companies I worked for created an Importance Table along with the measured dimensions (for supplier and on site manufacturing). We differenciated between Key and Critical Characterisitcs. Some characterisics were required to be >1.33 althe way to some being required > 2.00.

Excellent response AnnafromA2 and Prashant!

Can we now start the off topic responses about the world of Statistics?

(It starts with doing a little CPk and the next thing you know your out on the street trying to score 6s Black Belt Cert.)