Setting up SPC charts

[Amira1011]

Hello all,

Very lucky to be a part of this forum. I want to hear some feedback on my thought process here. Little bit of background about me, I am a recent college grad, with lot of theoretical knowledge about SPC but not a lot of practical knowledge.

So, I recently joined a company, where the SPC monitoring is out of place, control limits and spec limit are basically the same, not standardization in terms of test rule violations, etc. My task is to start from scratch and improve the basics i.e. setting up spec limits, control limits, etc.

Now, from my theoretical knowledge, when it comes to setting up your control limits, for example for X-bar chart, you use mean which would be x double bar and then you calculate your control limits based out of it (can be either static or dynamic). But my company that I work at, says that is not how you should do it at your practical work. They way you should do it is to take the target of specification and you use that value of target to calculate your control limits. While I understand that should be the ideal scenario where you'd want the mean of the data and your target to be equal, I think using the target value to calculate control limits may be misleading and would not help us to understand the actual process behavior.

Can someone please help me understand how it should be done in real practice? Thank you!

 

[Miner]

It depends on how easy it is to adjust the process mean. If the process mean is difficult to adjust to the target, then your approach is probably the way to go. However, if it is relatively easy to adjust the mean of the process then a target-centered control chart is very common.

 

[Golfman25]

Hello all,

Very lucky to be a part of this forum. I want to hear some feedback on my thought process here. Little bit of background about me, I am a recent college grad, with lot of theoretical knowledge about SPC but not a lot of practical knowledge.

So, I recently joined a company, where the SPC monitoring is out of place, control limits and spec limit are basically the same, not standardization in terms of test rule violations, etc. My task is to start from scratch and improve the basics i.e. setting up spec limits, control limits, etc.

Now, from my theoretical knowledge, when it comes to setting up your control limits, for example for X-bar chart, you use mean which would be x double bar and then you calculate your control limits based out of it (can be either static or dynamic). But my company that I work at, says that is not how you should do it at your practical work. They way you should do it is to take the target of specification and you use that value of target to calculate your control limits. While I understand that should be the ideal scenario where you'd want the mean of the data and your target to be equal, I think using the target value to calculate control limits may be misleading and would not help us to understand the actual process behavior.

Can someone please help me understand how it should be done in real practice? Thank you!

You're going to learn by doing. Be ready to try many different things to figure out what will work in your applications. They probably have done this over the years. You'll need to confirm whether that's the best way or another way is better. Good luck.

 

[Steve Prevette]

Definietely there is the concept of the "Voice of the Process" (SPC model) versus the 'Voice of the Customer" (the specs). A very easy (and inexpensive) read is Don Wheeler's Understanding Variation, the Key to Managing Chaos. Also, I have a lot of SPC materials posted here Free - Steve Prevette's Statistical Process Control (SPC) "Library"

Dr. Deming's Red Bead Experiment is a good convincer of skeptics. I do have a version here ->

 

[Bev D]

what Miner said - but if it isn’t easy to center the process then NO, NO, NO. Don’t confuse product conformance to specs with process control. Please start reading Donald Wheeler. (SPCPress.com - it’s free).

also if you can center the process easily then you probably have rational subgrouping problems…read Wheeler