Using SPC Control Charts to Reduce AQL Sample Size

T

Trisampson

Hi all,

Our company (a medical component manufacturer) utilizes "In-Process" inspections during the manufacturing process as well as an AQL lot acceptance ("Final") prior to shipping. We would like to reduce, or eliminate the AQL sample size based on the In-Process data. If we are using SPC, are there rules or references available to guide us in reducing the AQL sample size? One route is to use process capability statistics (such as a > 1.33 Cpk reduces the AQL index to 2.5).

Has anyone any recommendations on this?

Thanks in advance...
 
J

John Nabors - 2009

Trisampson-

Hi all,

One route is to use process capability statistics (such as a > 1.33 Cpk reduces the AQL index to 2.5).

Has anyone any recommendations on this?

Thanks in advance...

This is just one knuckehead's :2cents:, but I would be very skittish about this. It would be one thing if you were making Barbie clothes or birthday candles but in medical I would be very leery of loosening your AQL's.

What does your customer(s) think of this?

Regards -John
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
Check your contracts - most customers specify a required final inspection AQL prior to shipping. However, proper SPC will help you pass those inspections, and minimize any sorts, which are a bigger drain on the till.:cool:
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Hi all,

Our company (a medical component manufacturer) utilizes "In-Process" inspections during the manufacturing process as well as an AQL lot acceptance ("Final") prior to shipping. We would like to reduce, or eliminate the AQL sample size based on the In-Process data. If we are using SPC, are there rules or references available to guide us in reducing the AQL sample size? One route is to use process capability statistics (such as a > 1.33 Cpk reduces the AQL index to 2.5).

Has anyone any recommendations on this?

Thanks in advance...

One thing to consider is holding the AQL the same, but checking to see if you can drop from Normal to Reduced sampling.
 
Q

qualitytrec

One thing to consider is holding the AQL the same, but checking to see if you can drop from Normal to Reduced sampling.

I agree. Use the switching Procedures found in ANSI/ASQ Z1.4-2003. If things go poorly then you increase inspections if they go well you loosen them.

8. Normal, Tightened and Reduced Inspection

8.1 Initiation of Inspection.
Normal inspection will be used at the start of inspection unless otherwise directed by the responsible authority.

8.2 Continuation of Inspection. Normal, tightened or reduced inspection shall continue unchanged on successive lots or batches except where the switching procedures given below require change.

8.3 Switching Procedures.

8.3.1 Normal to Tightened.
When normal inspection is in effect, tightened inspection shall be instituted when 2 out of 5 or fewer consecutive lots or batches have been non-acceptable on original inspection (i.e., ignoring resubmitted lots or batches for this procedure).

8.3.2 Tightened to Normal. When tightened inspection is in effect, normal inspection shall be instituted when 5 consecutive lots or batches have been considered acceptable on original inspection.

8.3.3 Normal to Reduced. When normal inspection is in effect, reduced inspection shall be instituted providing that all of the following conditions are satisfied.

a.The precedinng 10 lots or batches (or more, as indicated by the note to Table VIII) have been on normal inspection and all have been accepted on original inspection; and

b.The total number of nonconforming units (or nonconformities) in the samples from the proceding 10 lots or batches (or such other number as was used for condition "a" above) is equal to or less than the acceptable limit number given in Table VIII (see 8.5). If double or multiple sampling is in use, all samples inspected should be included, not "first" samples only; and

c.Production is at a steady rate; and

d.Reduced inspection is considered desirable by the responsible authority.​

8.3.4 Reduced to Normal. When reduced inspection is in effect, normal inspection shall be instituted if any of the following occur on original inspection:

a.A lot or batch is rejected; or

b.A lot or batch is considered accptable under the procedures for reduced inspection given in 10.1.4; or

c.Production becomes irregular or delayed; or

d.Other conditions warrant that normal inspections shall be instituted.​

Hope this is helpful,
Mark
 
N

New to Quality

Hi! :bigwave:

I am new to Elsmar and I know this thread is a little old, but I am in a very similar situation at my company. We use a c=0 AQL with no heightened or reduced sampling plan for final inspection and typically our customers do not require a specific inspection process. We also do in-process inspections and are starting to collect the data electronically (yay!). The program we have can report the data in many ways, including process capability and control charts. We have found several processes with very high capability, but Quality does not know how to adjust inspection to this.

In the simplest terms, what I want to do is use the in-process inspection information and overall process capability to drive final inspection sampling in an easy to follow way. However, I'm a little stuck in the implementation of this. :confused:

Can I use a heightened/reduced sample size AQL plan using process data to switch instead of acceptance history? Or should I switch AQLs based on the process data (much easier since you only need one chart, but I don't think the acceptable quality level is really changing)? In any case, I am looking for a solution that has clear rules.

Thanks in advance if you can help me!
 

bobdoering

Stop X-bar/R Madness!!
Trusted Information Resource
I am new to Elsmar and I know this thread is a little old, but I am in a very similar situation at my company.

Welcome to the Cove! :D

Before we can help, we need some idea of what kind of processes you are trying to control. It helps us formulate more accurate response to your specific situation - and each situation is different - this is not a rubber stamp issue!

Another question, what kind of final inspection failure rate do you have now, and what is its affect? If you have significant reject rate, the first step of process control will be to get that first pass through rate to a very high percentage.
 
Last edited:
S

statuser

Hi, I read the statement "if Cpk> 1,33 then the AQL reduces to 2,5%." :confused:Can you explain how you get to this. I would expect that a Cpk has only 66 pppm defective which would be an AQL of 0,0066%

By the way does anybody knows how to design a sampling plan to guarantee a Cpk of 1,33, rather than quaranteeing an AQL of say 1%?

:thanx:
 
N

New to Quality

Hi again,

Thanks for the quick reply and sorry for the delayed response! The processes I am trying to control are primarily laser welding and cutting.

To start with a simple case: I have a process that is in control and has a process capability of 2.11 (measured over three lots). Over the past two years, we have rarely had any failures for this part and root causes have been addressed for these. Quality has continued to inspect at a .65 AQL since the process was developed regardless of process performance or inspection history.

Here are the paths I see for a case like this (and my thoughts on each)
1. Move to a higher AQL - this achieves the goal by reducing the sample size, but seems to be false since the accpetable quality level isn't changing from the customer perspective (although customer does not specify AQL and .65 was chosen internally). Also, I'm not sure what justification I would need for this...cpk, lot acceptance history?
2. Use a reduced sampling plan for AQL - this seems ideal because it maintains the AQL, but allows us to reduce sample size. My only issue with this is that it is only based on lot history not process performance and I am trying to encourage more emphasis on quality throughout the process instead of just a check at the end.
3. Make a new sampling plan - I want something that is statistically valid and easy to use. I'm thinking that quality could receive a lot of parts and a report with the control chart from the lot and the cpk.
- If there were any special causes of variation with a data point out of spec, quality would inspect 100%.
- If there were special causes of variation within spec limits, quality would inspect at a heightened level which they could look up on a simple chart.
- If the process is in control and the cpk is < 1.33 and > 1, inspect at normal level
- If process is in control and cpkk >1.33, inspect at reduced level.

Has anyone ever used a plan like I outlined in option 3? Is it valid to base final inspection sample size on an individual lot process performance?

Thanks!
 
Top Bottom