I-MR SPC used for Inspection

A

a team

Background to the question...I was at a company that is has quality auditors going from cell-to-cell for first piece and last piece inspections. The results were recored in an SPC software to create an I-MR. The quality auditors were instructed not to record any parts that failed the FP check...only record the one part the first good piece. (the approach is flawed) The auditor only records the first piece and last piece for a run for a total of two pieces per part off.

The question is this...has anyone else ran into or records only first piece checks in an I-MR. If so, do you feel this effective?
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Unless there is a specific belief as to why the first part and last part are critical, I would much prefer to see a random sample of the run. But (assuming that you change so ALL data are recorded as found), I don't see why an I-MR of the first and last piece would not have at least some value, assuming that you are not in a tool-wear scenario.
 

Bev D

Heretical Statistician
Leader
Super Moderator
I see this type of behavior (only recording the passing results) quite frequently in many companies. While well intended (the operator inspector or supervisor is focused on the need to record values representing what was shipped or accepted) it is wrong. Aside from the obvious censoring of data, There is much more value in the failing results than in the passing results.
 

Miner

Forum Moderator
Leader
Admin
Unless there is a specific belief as to why the first part and last part are critical, I would much prefer to see a random sample of the run. But (assuming that you change so ALL data are recorded as found), I don't see why an I-MR of the first and last piece would not have at least some value, assuming that you are not in a tool-wear scenario.
I agree totally with BevD's comments. To address Steve's comment: If the variation in a given process is primarily setup dependent, such as in a stamping operation, the first piece after the setup has concluded is an excellent indicator of the entire production run. If the major source of variation is not setup, then Steve's comment is dead on.
 
M

Matt33

Collecting data on the first and last part might be acceptable if only three parts were made. But if 3,000 parts were made, does the data obtained from the first and last adequately represent the rest of the data?
If ?Yes? is the answer, can this be proved?
If ?No? is the answer, then it seems logical to collect more data throughout the process.
 
D

Darius

Collecting data on the first and last part might be acceptable if only three parts were made. But if 3,000 parts were made, does the data obtained from the first and last adequately represent the rest of the data?
If “Yes” is the answer, can this be proved?
If “No” is the answer, then it seems logical to collect more data throughout the process.

The first part seem to be a critical point because is a change of setup of operation, :2cents:IMHO in precontrol is recommended not to take the first items because of that (you have a different condition, and you can't chart different conditions), and one can say the same for the SPC methodology, normaly you are not starting the machine.

As stated before on this tread is better to take samples randomly, but you can obtain the control limits for one case and the other (the current procedure against the randomized sampling) and check if they are different, so it can be proved (that they are different, not that it need more samples, but you can do the same for more sampling).
 

Bev D

Heretical Statistician
Leader
Super Moderator
Collecting data on the first and last part might be acceptable if only three parts were made. But if 3,000 parts were made, does the data obtained from the first and last adequately represent the rest of the data?
If ?Yes? is the answer, can this be proved?
If ?No? is the answer, then it seems logical to collect more data throughout the process.

it depends on the process and part/material

there are processes where there is no substantial change from start to finish except when a shift occurs and that shift is typically sustained. and yes it can and has been proven. htat is what capability studies are for. (real ones not just the 'check the box' Cpk study)

of course there are processes where it absolutely doesn't make sense to sample only the first and last piece.

As always the actual physics of the situation rule.
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
I agree totally with BevD's comments. To address Steve's comment: If the variation in a given process is primarily setup dependent, such as in a stamping operation, the first piece after the setup has concluded is an excellent indicator of the entire production run. If the major source of variation is not setup, then Steve's comment is dead on.

Usually last piece is there to pick up anything that might have occured after set-up, such as tool chipout, etc. Better to know now....
 
Top Bottom