Need journal and reference of abnormal bar chart SPC

gitabimbi

Registered
Hi all, I work in a die casting company. My company implement SPC in our process.
But, we only use 2 rules of abnormal graph.
rules 1 : ( Point out of control limit )
rules 2 : ( 4 out of 5 point in a row on same side of centerline )

I want to explain to management that we should use all kinds of abnormal charts. I got this knowledge when in my old company. but I do not have a qualified journal as a reference to management.

please share with me if you have the journal. Thank you..
 

AMIT BALLAL

Super Moderator
You can refer to AIAG's SPC 2nd edition manual / other manuals for the same.
Following is the snapshot of the rules for identifying special causes as per the AIAG's SPC Manual:
1627032836841.png

However, we won't be able to share the document here since it is copyrighted.
 

Miner

Forum Moderator
Leader
Admin
The rules that you use should correspond to the needs of your process. The more that you use, the greater the risk of a false alarm.

  1. Rule 1 should always be used.
  2. Rule 2 should work for most processes
  3. Rule 3 should only be used if your process is subject to trends
  4. Rule 4 usually indicates a problem with mixing process streams, and is not necessary once this problem is corrected
  5. Rules 5 & 6 are intended to detect smaller process shifts. Use these if your capability is marginal, and the process must be tightly controlled. If you have high capability, these are not necessary.
  6. Rules 7 & 8 are diagnostic tests to verify that the rational subgroups are correct. Once this has been verified, these should be dropped.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Remember, the point of the rules (usually the Western Electric Rules) is to find non-random behaviors, and Shewhart Charts specifically assume that the process is totally random unless some unusual special cause action is taking place causing the process to become out of control. It is critical to realize that your process must be expected to be random for it to work. However, some processes are not random, they are functions - which is why they taught you functions in school. Functions are neither random nor independent, but may be fully in control (if expected.) Trends are expected in functions. So, before locking your sights into forcing your process to meet these rules, make sure the process has random and independent output. If not, use some other statistical approach to control the process. Otherwise, you may be driven to overcontrol your process...which is bad.
 
Top Bottom