Effective SQC in a Manufacturing Plant with JMP?

[#1]

I am old school and believe that SQC is most effective when a physical control chart is in the hands of the operators of a plant.  In this way, the operator documents the exceptional variation, the causes for the variation are therefore captured, and improvement can commence from the learning.

Has JMP solved the problem with implementing SQC electronically that still enables capturing (and learning from) assignable cause variation?  I would prefer to use JMP for broad SQC implementation as my company has many JMP licenses.

[#2]

@toddsedwards this approach works well when there are a limited number of critical quality attributes or electronic collection of data is difficult. The big challenge is when there are more than a few CQAs or if they're correlated. That's where modern approaches, such as the one found in the Model Driven Multivariate Control Chart platform are helpful. JMP provides solutions (in certain contexts) of implementing SQC electronically. Given the breadth of SQC, I'm not sure I can provide a better answer without more context.

[#3]

Thanks for the reply. The context is receipt of a single continuous variable measurement every 4 to 8 hours.  Old school, and very effective, way is to plot the result on a paper control chart.  Document and learn from the exceptional variation.

Folks today do not, or refuse, to plot results manually.  I don't need model driven control charts for the application that I am seeking.  I simply need a way to replicate control charts in the way that Shewhart intended...on a paper control chart.  The operator can enter the data manually, no need for fancy data connections or scripting.  I am simply not aware of how to do this in JMP in real time in a manufacturing environment that captures Shewhart's original intent...that we learn from non-random variation and continuously improve.

[#4]

First welcome to the community.  There are quite a few really smart folks (present company excluded) here with differing thoughts and ideas.  I'll offer my thoughts:

First, I do agree getting folks involved in acquiring and analyzing data has many benefits.

1. Use of sampling is incredibly powerful, and analysis depends completely on how the data is collected.  Hopefully you are aware of the original intentions of Shewhart in using control charts.  Shewhart suggests using rational subgrouping ideas to determine subgrouping strategy.  The subgroups provide a basis of comparison.  The variation within subgroup is a function of the factors (x's) that change at that frequency. The range chart answers the question "is the variation within subgroup, a function of the x's changing within subgroup, consistent/stable?" The sampling frequency is also determined so as to capture the effect of x's changing at the selected sampling frequency.  The Y-bar (X-bar) chart compares the sources of variation within subgroup (control limits) to the sources of variation between subgroups to determine which source has the greatest effect.  These are not intended to be monitoring tools, but to answer specific questions as to where to focus subsequent work and and the nature of the investigation. (e.g., special causes that might make it easier to see effects at particular points in time).

2. JMP is more than capable of providing good graphical analysis of the data (e.g., variability plots or graph builder) as well as appropriate control charts, though the Shewhart control charts have made their way down the hierarchy to Legacy Control Charts.

3. I'm not sure what you mean by "implementing SQC electronically".  There are ways to get the software to automatically bring data into the software (and plenty of JSL scripts can be written to make this efficient)...But is this really what you want?  Use the charts to analyze your sampling plan and answer the questions you pose. Then change your sampling plan and repeat....

“The engineer who is successful in dividing his data initially into rational subgroups based on rational theories is therefore inherently better off in the long run. . .”

                                                                              Shewhart

[#5]

My needs are, conceptually, much simpler.  I need to get the control chart into the hands of the process operators who run the manufacturing plant.  I need them to document / annotate the continuous data in real time so that we all learn from the non-random variation that will be documented.

The control chart type is the simplest: Two-point moving range.  That single data point comes in every 4 to 8 hours. 

It is not apparent to me that JMP has this ability, which is why I am seeking help.

[#6]

Maybe I'm missing something, but if you want paper control charts at Gemba (Lean term for 'where the work gets done') and it sounds like you don't have a large number of critical to process variables, here's my three cents:

1. Set up a Lean Morning Market board with the control charts in paper form as one of the board elements. Make one of the operators the person that creates the paper charts on some cyclical interval...with enough blank x axis entries in the JMP data table so you have some blank space to hand plot future observations. Then part of Standard Work is as each data point becomes available, the designated operator hand plots the data, evaluate for potential action, or better yet, unless immediate action is called for, make that evaluation conversation part of the Morning Market meeting.

2. Then as time rolls along, the data can be updated in JMP, again on a cyclical basis. Once updated in JMP...reprint the chart(s). Post on the Morning Market board...return to step 1.

We did this sort of thing alot during my tenure at Eastman Kodak Company. Granted there needs to be some Lean Thinking infrastructure around the actual paper charts...like integrating Morning Market into the operations, and all the associated standard work around what happens based on the conversations that occur in the Morning Market...otherwise the paper chart exercise will just become one more thing that happens but nobody does anything about it.

[#7]

Sorry, but your interpretation of how control charts are used is conceptually different than mine.  I use then aggressively to separate and assign sources of variation and assess stability.  You are using them by reacting to out-of-control points.  I believe you are using IMR charts (no rational subgroups). 

If I understand the issue, you can't get them to write on paper, but you can get them to enter data into a confuser?  Create a JMP data table for them to enter the data when they get it.  Teach them to create the chart in JMP and then how to analyze it.

[#8]

The subgroups are rational, though n = 1.  So, yes, IMR chart.

We are using control charts in the same way, I think.  However, we do not react to normal variation.  We use the control chart to separate normal variation from exceptional (non-random) variation.  The aim is to get the organization to document causes of exceptional variation so that these sources of variation can be effectively eliminated.

Data visualization, like a paper control chart in real-time, is very effective in motivating the desired organizational response (as outlined above).  I am searching for a way to emulate a paper chart in JMP.  It seems that JMP does not have this functionality and that I may need to consider different software unless someone in this community has solved this problem in JMP.

[#9]

I think you can come pretty close with JMP. Create a JMP table with some of the data in it. Now create your control chart and leave it open. Now add an additional row to the data table. The control chart will automatically show the additional point. Isn't that what you would do on a paper chart?

You can even have a blank data table and create the empty control chart. Every time you add a point to the data table, it will update the chart.

[#10]

Thanks.  My issue is text annotation (free form) of the assignable cause of variation of the point in JMP.  Any ideas on how to do that?