Questions - AQL & RQL

[MedtechQuality]

1) For AQL = 1% and a lot size of 1000 units, the sample size and acceptance number can be determined as follows: AQL = 1% Sample Size =80 Acceptance Number (Ac) = 2 Rejection Number (Re) = 3 Based on the above details, I start my inspection and declare the lot rejected as soon as I find 3 non-conforming units before even going through the entire sample size of 80. Why would I even bother to look into RQL?

2) For AQL = 1% and a lot size of 1000 units, the sample size and acceptance number can be determined as follows: AQL = 1% Sample Size =80 Acceptance Number (Ac) = 2 Rejection Number (Re) = 3 Based on the above details, I start my inspection and I find only 2 non-conforming. So, I declare the lot accepted. But, I know there are non-conforming units in the lot. How can I accept this lot as it will affect the downstream process and the end client?

3) How does an engineer select an AQL value? Is this based on the risk (severity)? Just like how we select the sample size for TMV by determining the severity of what is being measured from the PFMEA --> select the appropriate confidence and reliability number --> check the SOP for sample size.

All the above questions are raised with respect to a medical device manufacturing company.

Thanks in advance.

 

[Bev D]

First try a search of AQL RQL and my user name. (Upper right hand corner magnifying glass with search. We have answered this many many times so you will no doubt find this helpful.

Second your first question doesn’t make sense can you rephrase/clarify?

Your second question is answered simply by your use of the AQL. AQL means Acceptable Quality Level. Literally. If you state a 1% AQL then you are declaring that a lot with 1% defective material is acceptable. If your Customer accepts your AQL then they will be accepting that they may get 1% defective lots.

IF you are required or you desire to ship no defects then you must perform 100% inspection or test with a method that has very excellent repeatability & reproducibility and a high accuracy. Excellent process controls to prevent defects are also a good idea

As for your third question: yes the AQL and RQL come from the severity. BUT REMEMBER that AQL is the ACCEPTABLE quality level and RQL is the unnacptable (rejectable) quality level…your Customer cares about the RQL.

 

[Apex Hao]

First try a search of AQL RQL and my user name. (Upper right hand corner magnifying glass with search. We have answered this many many times so you will no doubt find this helpful.

Second your first question doesn’t make sense can you rephrase/clarify?

Your second question is answered simply by your use of the AQL. AQL means Acceptable Quality Level. Literally. If you state a 1% AQL then you are declaring that a lot with 1% defective material is acceptable. If your Customer accepts your AQL then they will be accepting that they may get 1% defective lots.

IF you are required or you desire to ship no defects then you must perform 100% inspection or test with a method that has very excellent repeatability & reproducibility and a high accuracy. Excellent process controls to prevent defects are also a good idea

As for your third question: yes the AQL and RQL come from the severity. BUT REMEMBER that AQL is the ACCEPTABLE quality level and RQL is the unnacptable (rejectable) quality level…your Customer cares about the RQL.

hi Bev, let's say one particular failure is of major risk, and thus the process control has been thoroughly verified and validated.

Under such situation, if my organization still wish to conduct sampling check on finished product (as proof of confidence), how would you recommend to proceed? The dilemma is that, for a major risk, our required AQL is 0.65. For lot size of 1000, we need to sample 80 units. This seems excessive since we already have a robust process control, but we can't just adjust the AQL to lower the sample size.

Appreciate your feedback.

 

[Bev D]

You are correct about the sample size - you just don’t like it.

Reducing the sample size is a waste of time in inspection as you are unlikely to ever catch a defect rate that you don’t want.

The only other thing you can do is to sample 3-5 units at every process change point (like SPC).

 

[Apex Hao]

You are correct about the sample size - you just don’t like it.

Reducing the sample size is a waste of time in inspection as you are unlikely to ever catch a defect rate that you don’t want.

The only other thing you can do is to sample 3-5 units at every process change point (like SPC).

Just to clarify, so this initiative (sampling check) on top of existing process control which is robust, does nothing to help reducing effort in sampling?

We can do SPC. Is it valid to sum up all those quantity from SPC in a single lot, as the final sampling number in Z1.4 sampling plan? Our customer is requesting AQL, and I doubt doing SPC can fully replace it.

 

[Bev D]

Just to clarify, so this initiative (sampling check) on top of existing process control which is robust, does nothing to help reducing effort in sampling?

We can do SPC. Is it valid to sum up all those quantity from SPC in a single lot, as the final sampling number in Z1.4 sampling plan? Our customer is requesting AQL, and I doubt doing SPC can fully replace it.

IF you have a robust control plan, the only reduction in INSPECTION sampling is to zero. Anything more than zero but less than the AQL* is a waste and does nothing to ‘improve confidence’. Remember that AQL/RAL inspection sampling is based on the ‘confidence’ you have in the stated defect rate. If you reduce the sample size you increase the allowable defect rate. That’s just a fact.

NO it really isn’t valid to combine the smaller samples to add up to the AQL level as acceptance sampling is intended to be RANDOM and any SPC or other change point sampling is the exact opposite of random. It is however, actually more powerful than acceptance sampling of a finished lot for all but ‘random’ defects. But your process may not have ‘random’ defects as these are typically due to isolated human error…

If your Customer is requesting AQL sampling perhaps they might accept a combination of small samples. Or SPC samples…you would have to ask them.

For very serious severe defects 100% inspection is the most reliable approach. A ‘small’ AQL is really just whistling in the graveyard in an attempt to avoid 100% inspection. Also most peopel are still not recognizing that AQL means ACCEPTABLE DEFECT RATE. So if you ahve an AQL 0.65 than if htat defect relate is present you will almost always ACCEPT it and SHIP IT.

Statistics can’t do magical things….