Jump to content

  • Quick Navigation
Photo

How to set UCL & LCL when creating new specifications?


  • You cannot start a new topic
  • Please log in to reply
9 replies to this topic

#1 mohamed ahmed yusuf

mohamed ahmed yusuf

    Grade - MIFSQN

  • IFSQN Member
  • 162 posts
  • 37 thanks
14
Good

  • Egypt
    Egypt
  • Gender:Male
  • Interests:Reading, Listening music, handball , volleyball, food safety, drawing , R&D, Photoshop

Posted 05 May 2018 - 08:14 AM

Hello, 

I'm now working on creating a new specs and i'm facing a difficulty in how to set the upper control limit and lower control limit. 

the data is non normal and have many outliers , my question is about how to set the UCL & LCL and also if i remove the outliers will this affect on the data ? 

 

Thanks for your help

Regards 

Mohamed 


M.Yusuf


#2 Charles.C

Charles.C

    Grade - FIFSQN

  • IFSQN Moderator
  • 18,453 posts
  • 5165 thanks
1,154
Excellent

  • Earth
    Earth
  • Gender:Male
  • Interests:SF
    TV
    Movies

Posted 05 May 2018 - 10:23 AM

Hello, 

I'm now working on creating a new specs and i'm facing a difficulty in how to set the upper control limit and lower control limit. 

the data is non normal and have many outliers , my question is about how to set the UCL & LCL and also if i remove the outliers will this affect on the data ? 

 

Thanks for your help

Regards 

Mohamed 

 

Hi mohamed,

 

A little context needed.

 

Are you referring to Statistical Process Control ?

 

If so, for what product, process, control parameter ?

 

If you mean a Product Specification, need some more similar context also.


Kind Regards,

 

Charles.C


Thanked by 1 Member:

#3 Mohamed.reda

Mohamed.reda

    Grade - AIFSQN

  • IFSQN Associate
  • 39 posts
  • 2 thanks
0
Neutral

  • Egypt
    Egypt

Posted 06 May 2018 - 10:02 PM

wellcome mohamed please check attchments may it's help

Attached Files



Thanked by 2 Members:

#4 mohamed ahmed yusuf

mohamed ahmed yusuf

    Grade - MIFSQN

  • IFSQN Member
  • 162 posts
  • 37 thanks
14
Good

  • Egypt
    Egypt
  • Gender:Male
  • Interests:Reading, Listening music, handball , volleyball, food safety, drawing , R&D, Photoshop

Posted 07 May 2018 - 04:55 PM

Hi mohamed,

 

A little context needed.

 

Are you referring to Statistical Process Control ?

 

If so, for what product, process, control parameter ?

 

If you mean a Product Specification, need some more similar context also

Sorry Charles for my late reply  :smile: ,

yes it is kind of SPC and i found a solution for the problem finally, really appreciate your help. 


M.Yusuf


#5 mohamed ahmed yusuf

mohamed ahmed yusuf

    Grade - MIFSQN

  • IFSQN Member
  • 162 posts
  • 37 thanks
14
Good

  • Egypt
    Egypt
  • Gender:Male
  • Interests:Reading, Listening music, handball , volleyball, food safety, drawing , R&D, Photoshop

Posted 07 May 2018 - 04:57 PM

wellcome mohamed please check attchments may it's help

Thanks Mohamed for your help , i will check it . 


M.Yusuf


#6 frawat

frawat

    Grade - Active

  • IFSQN Associate
  • 17 posts
  • 3 thanks
0
Neutral

  • Peru
    Peru
  • Gender:Male
  • Location:Peru
  • Interests:Quality Management
    Food Safety
    Statistical Thinking
    Six Sigma

Posted 08 May 2018 - 06:31 PM

Dear Mohamed.

 

You said you already had a solution but your subject is fascinating, maybe you coud share a little bit of this solution?

 

I think Charles.C questions are very pertinent.

 

Please note you need a predictable process to be able to comply with specifications, otherwise your would be in the "brink of chaos" as Dr. Wheeler says.

You need to eliminate special cause effects and standardize your process, and hopefully that should reduce variation. Then set a good process target.

 

I would only use UCL and LCL for checking if the process is predictable and keep a watch on special causes (SPC).

 

To check against specification maybe Cpk and Ppk charts could be useful.

 

Kind regards,

Francis



Thanked by 2 Members:

#7 frawat

frawat

    Grade - Active

  • IFSQN Associate
  • 17 posts
  • 3 thanks
0
Neutral

  • Peru
    Peru
  • Gender:Male
  • Location:Peru
  • Interests:Quality Management
    Food Safety
    Statistical Thinking
    Six Sigma

Posted 08 May 2018 - 06:41 PM

Sorry, you can also use USL and LSL to plot graphs on defects (p or np) or any measurable variable, but this would not be SPC.

So you could graph p or np against spec tolerances, or graph p or np or x against UCL and LCL in which case you are applying SPC.



Thanked by 1 Member:

#8 Charles.C

Charles.C

    Grade - FIFSQN

  • IFSQN Moderator
  • 18,453 posts
  • 5165 thanks
1,154
Excellent

  • Earth
    Earth
  • Gender:Male
  • Interests:SF
    TV
    Movies

Posted 08 May 2018 - 10:34 PM

Sorry Charles for my late reply  :smile: ,

yes it is kind of SPC and i found a solution for the problem finally, really appreciate your help. 

 

As per post 4 you typically need a stable process to implement spc formulae for 3sigma control bands, etc.

 

Sometimes non-normal data is correctable by changing to logs.

 

I once tried to do it for production micro. data but the measurements simply had too much intrinsic scatter so no usable mean value.


Kind Regards,

 

Charles.C


Thanked by 1 Member:

#9 Steve Gruler

Steve Gruler

    Grade - Active

  • IFSQN Active
  • 1 posts
  • 1 thanks
0
Neutral

  • United States
    United States

Posted 09 May 2018 - 04:33 PM

Mohamed

 

Spec limits are where product failure occurs once breached. Control limits are based on your production process capability.

 

Regards 

 

Steve

 

 



Thanked by 1 Member:

#10 mohamed ahmed yusuf

mohamed ahmed yusuf

    Grade - MIFSQN

  • IFSQN Member
  • 162 posts
  • 37 thanks
14
Good

  • Egypt
    Egypt
  • Gender:Male
  • Interests:Reading, Listening music, handball , volleyball, food safety, drawing , R&D, Photoshop

Posted 14 May 2018 - 03:56 PM

Hello, 

Sorry for my late reply , the site here is somehow difficult to use.

I checked all your comments and i'm grateful for you all. 

I Contacted with my instructor which taught me Lean six sigma green belt and he told me , you should do the following steps, i'm using Minitab program for analyzing data: 

1. should checking the normality data by checking the p value. if okay then you can run a control chart and seeing what will be the UCL and LCL. 

2. If data weren't normal then you have to make an outlier test which will determine the outliers and then,

3. remove it , and make again normality test if the data turned to be normal then you can use it, if it still abnormal data then you shouldn't remove any outlier from the data. 

4. Then after checking it , run the control charts and check whether the UCL & LCL are suitable for you and from the control chart there will be points out of control check these points.   

5. Then these UCL & LCL should be studied and see the capability of the process if it can applied. 

 

  • BTW for people who don't have minitab or another program it could be done by Excel also and this is a link for the video :

       

  1.  

 

That's all what i got and what i did , please share your comments. 

Apologize again for my late reply.

Regards 

Mohamed 


M.Yusuf





0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users