Hello,

I have been trying to calculate control limits (Xbar R charts) for 100 variables using Minitab. :bonk:

Variable types: 100
subgroup size: 5

All of this data is stored in excel columns in this way:

date | x1 | x2 | x3 | x4 | x5 | Variable type

Somebody knows how to perform limits calculation (Xbar, R) for each of the variables without performing "copy paste" each one of the 100 variable types?:confused:

Thanks,
Dave

It is not totally clear what you are requesting because you have to copy and paste from Excel into Minitab. If you mean how do you get the data from the 5 columns (1 column per measurement) into a format that Minitab can analyze, you need to use Minitab's Data, Stack Columns command. Then tell Minitab that you want to stack columns X1 through X5.

It is not totally clear what you are requesting because you have to copy and paste from Excel into Minitab. If you mean how do you get the data from the 5 columns (1 column per measurement) into a format that Minitab can analyze, you need to use Minitab's Data, Stack Columns command. Then tell Minitab that you want to stack columns X1 through X5.

Thanks for the reply but this is not what I meant. :)
I have a process with approximately 100 Jigs.
At each Jig I measure 5 points (X1 through X5)
and wish to calculate the limits for each of the 100 Jigs.
So basically I want to use the Xbar chart, in which each point is represented by an average of mentioned above 5 points - x1 through x5.

Ok, now the problematic part :rolleyes:
All of the data for all of the Jigs is present in one excel file in form:
date | x1 | x2 | x3 | x4 | x5 | Jig#

and I need to calculate the limits for each of the 100 Jigs.
I know that I need to copy paste the information into minitab, but how to perform the calculation so I will receive limits per each Jig?

Any body? :bigwave:

I'm not exactly clear as to what the OP means by limits?

BUT the Xbar R chart won't give him what he wants. The Xbar Chart will only calculate Statistical Control Limits for the 100 AVERAGES as a whole - a single uppper control limit and a single lower control limit. It will also calculate a single RANGE control limit for the acerage range of the 100 subgroups of 5 readings each.

It sounds as if the OP is asking how to use a screwdriver as a drill.

If the OP can clarify:

What is the function or purpose of the jigs?
What are the 5 readings? for example are they 5 different features on the jig? or are they 5 different parts processed on the jig?
What are the "limits" you want? for example: calibration limits for the jig? or is it for the product processed on teh jig?

What is the function or purpose of the jigs?

Mechanical assembly


What are the 5 readings? for example are they 5 different features on the jig? or are they 5 different parts processed on the jig?


5 reading of the same part - measurement of the same dimensions in 5 locations (these 5 points will represent the Jig)


What are the "limits" you want? for example: calibration limits for the jig? or is it for the product processed on teh jig?

Based on 5 points measured per part (those 5 points will represent the "Jig" status) I want to calculate the control limits.

Hope I clarified myself correctly this time :o

Do all 100 Jigs assemble produce same assembly or different assemblies? :confused:
It might be easy to provide part of the data and a few pictures with locations where data is being collected on Assemblies (to help understand from your explanation provided thus far.)
Sorry could not be of much help.

to clarify further: the jigs are used to assemble parts. You have taken one part assembled on each of teh 100 jigs. you measured the same feature on each part in 5 different locations on the feature.

is this correct?

Now what are you trying to control or monitor? the feature as assembled? (is the feature a result of assembly of sub parts on the jig?)

on the surface this doesn't sound like a rational subgrouping scheme. in general each jig would have it's own chart and one would subgroup by 5 parts - not 5 readings on a single part - for each jig. There are other approaches that may be appropriate but without more data and a better explanation of waht exactly you are trying to monitor or control we can't help you

to clarify further: the jigs are used to assemble parts. You have taken one part assembled on each of teh 100 jigs. you measured the same feature on each part in 5 different locations on the feature.

is this correct?

Now what are you trying to control or monitor? the feature as assembled? (is the feature a result of assembly of sub parts on the jig?)

on the surface this doesn't sound like a rational subgrouping scheme. in general each jig would have it's own chart and one would subgroup by 5 parts - not 5 readings on a single part - for each jig. There are other approaches that may be appropriate but without more data and a better explanation of waht exactly you are trying to monitor or control we can't help you

Bev,
I agree with your comments, hence request to get more clarification. Pictures are worth 1000 words.
Typically in assembly line a jig is used to assemble and create a feature that should be measured and controlled. Multiple operators using multiple jigs producing same part in an assembly cell type of situation can support using Xbar-R control chart for a CTQ feature for customer reporting.
But again I am not sure if I am in-line with what is being asked so any help can be provided:confused::confused:.

Bev,
I agree with your comments, hence request to get more clarification. Pictures are worth 1000 words.
Typically in assembly line a jig is used to assemble and create a feature that should be measured and controlled. Multiple operators using multiple jigs producing same part in an assembly cell type of situation can support using Xbar-R control chart for a CTQ feature for customer reporting.
But again I am not sure if I am in-line with what is being asked so any help can be provided:confused::confused:.

Hi Bev and Ashmot,

I am Attaching a sample of the SPC Data.
I agree with you that this case is not a typical "subgroup by 5 parts" - but 5 readings. The main reason is convenience and monitor.
In order to control those 5 readings I would need 5 charts (per Jig). But after evaluating the data, it was found that combining those readings into Xbar/R chart will fit the needs as well.
So basically reducing the monitor of 500 charts to 200.
My problem is how to calculate the control limits (not the equation) - performing the calculation for 100 Jigs with Minitab.

could any one help me know what is minitab?
Then I will solve this control limit puzzle.
I am an SPC buff.
:thanx:

v.j.brahmaiah

could any one help me know what is minitab?
Then I will solve this control limit puzzle.
I am an SPC buff.
:thanx:

v.j.brahmaiah

Minitab is a statistical software by which you can perform various analysis and calculations. Here is a link for the software website: http://www.minitab.com/

dkharif - I understand now what you want but we can't always get what we want. we must first understand our data - how does our process actually vary - then we can select the best control scheme.

In looking at your data teh first thing that stands out immediately is that the within piece variation (of the 5 readings per part) have relatively little variation compared to the part to part variation. Now it is impossible to determine at this point if the part to part variation is due to simple variation in the assembly process and the component parts OR if it due more to the individual jigs as you do not have data of multiple parts from the same jig. I fyou were to go forward with control limits based on a subgroup scheme of multiple readings per part constituting a subgroup, then you will have very narrow limits on the averages and almost every part will be outside the limits. (you have an irrational subgrouping scheme)

My advice would be to get data on 21 parts (7 sets of 3 sequential parts) from each of ten or so jigs and then let's look at that data. I'm sure it will show us which way to subgroup and that will lead us to the control limit calculations.

dkharif - I understand now what you want but we can't always get what we want. we must first understand our data - how does our process actually vary - then we can select the best control scheme.

In looking at your data teh first thing that stands out immediately is that the within piece variation (of the 5 readings per part) have relatively little variation compared to the part to part variation. Now it is impossible to determine at this point if the part to part variation is due to simple variation in the assembly process and the component parts OR if it due more to the individual jigs as you do not have data of multiple parts from the same jig. I fyou were to go forward with control limits based on a subgroup scheme of multiple readings per part constituting a subgroup, then you will have very narrow limits on the averages and almost every part will be outside the limits. (you have an irrational subgrouping scheme)

My advice would be to get data on 21 parts (7 sets of 3 sequential parts) from each of ten or so jigs and then let's look at that data. I'm sure it will show us which way to subgroup and that will lead us to the control limit calculations.


From my analysis, part to part variation is ,as you have mentioned, due to assembly process, but also due to the jigs. Each Jig behaves differently - some with high variation, and some with low variation and with high Cp and Cpk values.
This is the reason that I need to analyze/control each jig individually.
In my opinion, xbar/R chart for he 5 readings per part will be sufficient for analysis/control.

Im attaching Excel file with some of the results.

From my analysis, part to part variation is ,as you have mentioned, due to assembly process, but also due to the jigs. Each Jig behaves differently - some with high variation, and some with low variation and with high Cp and Cpk values.
This is the reason that I need to analyze/control each jig individually.
In my opinion, xbar/R chart for he 5 readings per part will be sufficient for analysis/control.

Im attaching Excel file with some of the results.

The variation in capability may indicate you may need to improve the jigs to allow the use of SPC charting. At a minimum, with as much variation as you say you have, you would need to track each jig individually - as if they were individual machines or process streams. The only way to not do it is to have jigs so consistent that there is not statistically significant difference between the jigs, that is the jig the part is in makes no real difference. As far as control, what would be the action taken if the dimension was out of control? Sort? Replace the jig? Are there adjustments? What is the end game for implementing the control charts?