# Process Capability Study using Variables Data (weight of the product)

A

#### arperez

Dear Ms. / Mr.

I need help for a study of process capability as a variable using the weight of the products.

I have taken 50 measurements every half hour in six subgroups.

Have you any idea how to do this study? Where can I find some document excel?

I waiting for their comments and recommendations.

Greetings.

Last edited by a moderator:

#### bobdoering

Trusted Information Resource
Re: Process capability study.

Ah....there is no "spreadsheet" to plug data in to get an understanding of the process capability. If you could post your data, we can review the results, going through the steps to analyze the data.

To accurately understand capability, it would be good to understand the process. What is the process? What are the sources of variation in the process? Can you prepare a Total Variance Equation? What is your gage R&R - how much of the variation in your data is attributable to the variation in weighing? Remember, the output of a process is not one distribution, but the sum of all distributions in the Total Variance Equation.

Staff member

#### bobdoering

Trusted Information Resource
If anyone can comment on the original poster's specific "...50 measurements every half hour in six subgroups..." situation it will be appreciated.
My comment was specific to that post. There is no way of knowing if that frequency is adequate, or if the data is correct for the process by simply churning it in a spreadsheet. We have no idea of the significance of the 6 subgroups. That will fall into a condition similar to the Minitab Crutch Syndrome.

The data needs to be looked at in time sequence, and it also needs to be evaluated for its underlying distribution(s). Process capability is process knowledge, not the only slightly meaningful calculations that yield Cp and Cpk. I would avoid directing the poster to a plug and chug Cp/Cpk spreadsheet to further encourage such bad habits.

The proper capability evaluation steps:

1. FMEA or Fishbone to evaluate the sources of variation
2. Total Variation Equation, as well as estimates of each variation factor's contribution and distribution.
3.Elimination of of possible non-process variation prior to the capability study (proper gage R&R, etc.)
4. Collection of time sequence data
5. Observation for any trending, etc. that would ensure the data is non-normal or evidence that the underlying process variation is truly normal.
6. Determination of the correct process distribution
7. Estimate (because they are ALL estimates based on the sample) of the process capability

Not sure of any spreadsheet that does all of this.

#### Stijloor

Staff member
Super Moderator
Yep! Great summation of actions required...

There's no "Instant Pudding." (Credit to Dr. W. Edwards Deming)

Stijloor.

#### Marc

##### Fully vaccinated are you?
Staff member
<snip> 1. FMEA or Fishbone to evaluate the sources of variation
2. Total Variation Equation, as well as estimates of each variation factor's contribution and distribution.
3.Elimination of of possible non-process variation prior to the capability study (proper gage R&R, etc.)
4. Collection of time sequence data
5. Observation for any trending, etc. that would ensure the data is non-normal or evidence that the underlying process variation is truly normal.
6. Determination of the correct process distribution
7. Estimate (because they are ALL estimates based on the sample) of the process capability <snip>
I think step 7 is what the original poster is asking about. Let's *assume* the original poster has done steps 1 through 6 and is ready to do a study (I think this is your step 7).

While I appreciate the details you provide of all of the required steps starting with the FMEA or Fishbone, there are a lot of people who come here for spreadsheets where you can plug in 'typical' data which is why I linked to the existing spreadsheets here. I am not aware of one which addresses 50 measurements every half hour in six subgroups, however.

I am not aware of any single spreadsheet, or any single software program, that does everything you list in an integrated fashion either, but there are spreadsheets and/or software programs which address one or more of the individual steps you have listed.

#### bobdoering

Trusted Information Resource
I am not aware of one which addresses 50 measurements every half hour in six subgroups, however.
It is still not clear if there is a language barrier, and that what he is trying to explain is that he has subgroups of six specimens taken every half hour for a total of fifty samples (which is pretty common). If he has 50 specimens taken every have hour for a total of 6 samples, it can still be done, but you need a table with d2 for 50 specimens (reference Statistical Quality Control, seventh edition, Grant and Leavenworth, ISBN 0-07-024162-7, Table C, page 717).

Several of your cited spreadsheets can use the X-bar-R charting to estimate the standard deviation for the capability calculation if it is known that the process variation is a be a normal distribution. Otherwise, the technique is suspect at best, or simply incorrect. So, we still have too little information to direct him to the most correct answer.

A

#### arperez

Dear friends.

I just uploaded a document in excel where I have the measurements so you can see the information. I explain the process where I'm trying to make the process capability study, is called "packaging", there we made the filling of aerosols, taking into account the specifications of our customers.

I have taken measurements corresponds to a aersol X, which is introduced inside a concentrate and a propellant. We want to test your ability using the variable weight. Because we know if our process meets the specifications for filling, as we asked the customer.

At the customer's specification, we are told that we must concentrate and 4.6 grams 5.0 grams of propellant, the container valve more weigh 6.8 grams and the total weight is 16.4 grams.

Measurements are of the concentrate supply them first and the second are the propellant.

I hope you understand me.

Greetings.

#### Attachments

• 34 KB Views: 266

#### Bev D

##### Heretical Statistician
Staff member
Super Moderator
You have a larger problem than which formula to enter in your spreadsheet. (marc has posted the appropriate threads for hwo to do this)

I've looked at your data and any of the standard calculations you make will give you poor results as you have something known as "chunky data".

Chunky data is where you have very little resolution in your measurement system compared with your process variation. In practice it means that you will have only 2-4 numerical values showing up over and over in your data.

For example your first data has only values of 4.6 or 4.7.

Chunky data results in gross overestimation of the standard deviation regardless of whether you use the range or standard deviation method. So when you do the calculation you will get a low Cpk/Ppk value. But your process is fully capable - at least across the time frame you sampled.

As always the best assessment of process capability is to plot your data in time sequence against the spec limits and think about what that means.