SPC System Requirement(s) BrainStorm

[hazwan2283]

Hi everyone, my name is Hazwan.
I was given a task by my superior recently where i need to list down what are the specific requirements a good SPC system should have.
So that i can give these requirements to my IT team where we will spend some time (long time) to develop internal in-house SPC system.
These are the requirements i can think of, please correct me and add if you see i'm missing some points.
1. The system should be able to retain data (measurement data and all other data) for at least 15 years (not IATF requirement but customer req).
2. System should be able to trigger email for at least all the 8 SPC WECO Rules to the respective person-in-charged.
3. System should be able to calculate and display the capability indices (Cpk/Cp/Ppk etc).
4. System should be able to display control limit, spec limits.
5. Should be able to generate some raw data report just in case if process engineers need them for analysis purpose.
6. Minimally should be able to plot variable chart (X bar R/S and I-MR) and also attribute chart (p/np/c/u).
7. Should be able to store some information related to the process (operator name, machine number etc).
8. Should able to store root cause and actions taken for the WECO/OOC/OOS incidents.
9. Should able to display special characteristics symbol (per IATF requirement).

I'm aware most of above so called 'requirements' are not really IATF requirement but anyway is there any specific requirement from IATF in terms of SPC system that i missed out or did not include?.

 

[Steve Prevette]

Well, I ran a SPC based dashboard for 20 years. You may want to consider some features of it I used. Steve Prevette's Statistical Process Control (SPC) "Library"

Also, an overview at

 

[Bev D]

I have lead the creation of home grown SPC systems for 40 years (back in the day before dinosaurs roamed the earth I partnered with a developer to create a SPC system that displayed on a VT 101 monitor. The graphs were vertical rather than horizontal but it worked really well).
I have a list of requirements too that I will try to post later. Steve’s are pretty great by the way…one thing I will correct is that spec limits are useless unless on an I, MR chart of individual values. DON’T put them on an Xbar, R chart.

 

[Miner]

Validate the correct calculation of the control limits.

See Dr. Wheeler's articles"

• Individual Charts Done Right and Wrong How does your software rate? Donald J. Wheeler
• The Right and Wrong Ways of Computing Limits How does your software measure up? Donald J. Wheeler

 

[hazwan2283]

Thank you very much @Miner , @Bev D and @Steve Prevette . Really helpful sharing. I watched the video Steve posted and also wondering what does Bev means when spec limit is useless for Xbar-R Chart?...can you kindly elaborate Bev?..

 

[Steve Prevette]

What Bev is referring to is if let's say the subgroup size is 4 in my Xbar chart. Let's say there was one item that exceeded a spec. On the individuals chart, I could see that that item exceeded a spec limit. However, on the Xbar chart, that result would be averaged with three other results and might not be seen as exceeding the spec. The control limits work, as they are set knowing what the subgroup size is. You might see the effect of the item that exceeded spec on the Range chart, but also may not be fully relieable in that manner in terms of realizing you had an out of spec item.

THe control limit on the individuals chart also gives you a good visual clue as to how capable the process is..

 

[hazwan2283]

What Bev is referring to is if let's say the subgroup size is 4 in my Xbar chart. Let's say there was one item that exceeded a spec. On the individuals chart, I could see that that item exceeded a spec limit. However, on the Xbar chart, that result would be averaged with three other results and might not be seen as exceeding the spec. The control limits work, as they are set knowing what the subgroup size is. You might see the effect of the item that exceeded spec on the Range chart, but also may not be fully relieable in that manner in terms of realizing you had an out of spec item.

THe control limit on the individuals chart also gives you a good visual clue as to how capable the process is..

That makes sense, thank you Steve for the example

 

[Bev D]

Another requirement set:
The software should NEVER continuously recalculate the limits.
The software should allow engineers to enter the control limits OR the software should allow engineers to calculate the limits within the system.
The software should maintain and plot old limits only on the time frame for which those limits were valid. The software should only plot the new limits for the current time frame

You would also benefit greatly from adding the ability to plot your data in multi-vari format with spec limits.

 

[MOester]

I would add - the x-axis should be able to display TIME. Not just readings. If you take readings, say, on Mon, Wed and Fri, but few readings on Tue, Thu, you should see clusters of more ports grouped on M-W-F. It is often important to plot the data with TIME instead of just as an ordinal.

 

[hazwan2283]

thj

Another requirement set:
The software should NEVER continuously recalculate the limits.
The software should allow engineers to enter the control limits OR the software should allow engineers to calculate the limits within the system.
The software should maintain and plot old limits only on the time frame for which those limits were valid. The software should only plot the new limits for the current time frame

You would also benefit greatly from adding the ability to plot your data in multi-vari format with spec limits.

thank you very much Bev