1. Hello and Welcome to The Quality Forum Online...Continuing in the spirit of People Helping People !
    Dismiss Notice
Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

ANSI/ASQ Z1.9 Sampling Standard

Discussion in 'Sampling, Standards and Inspection' started by Mike S, Jan 11, 2016.

  1. Mike S

    Mike S Member

    Joined:
    Jan 11, 2016
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Hey there. I was a very infrequent poster at Elsmar, but I found it to be the best resource for QC questions that crop up in my job from time to time. I was really bummed to see that it shut down. Foudn out on the same day that I found out David Bowie died. Double whammy...

    II do have a few questions regarding ANSI/ASQ Z1.9, but I wasn't sure where to put them, so I guess I'll put the question here and you guys tell me where it goes (or help me here if you wish).

    I have an acceptance sampling plan by variables in use now. Everything was going great... until I had a lot come back as non-conforming. Looking at the data, out of 150 samples (lot size = 80,000 pcs), there was a single outlier that threw off the mean and standard deviation enough that it meant a lot non-conformance.

    I really don't want to scrap an entire 80,000 unit lot based on a single outlier. I also don't want to accept a lot if there's even 1% bad parts. AQL = 0.65, Normal, General Inspection Level II (Code = M). What is the general approach to investigating and re-submitting the lot? Do I move to a Tightened Inspection? How does that work? Do I change the AQL? Can you even re-submit the lot like this at all?

    Any help appreciated. Thanks.
     
  2. MarkMeer

    MarkMeer Well-Known Member

    Joined:
    Dec 3, 2015
    Messages:
    138
    Likes Received:
    62
    Trophy Points:
    27
    I'd imagine it depends on:
    a) the nature of the product (risks involved)
    b) the nature of your system (how much control you have)

    If the risk of accepting non-conforming product is really high, or if you don't have much control of the system (e.g. perhaps upper management, or customers push strict adherence to ANSI/ASQ Z1.9), then your hands may be tied.

    If, on the other hand, you can affect the system use this as an opportunity to modify the system, or make allowances for acceptance of certain product, under certain conditions, base on your own risk-assessment. Because, at the end of the day, it boils down to: what level of confidence do YOU have?

    If you suspect that there are a tiny number of outliers, one suggestion (again, assuming no strict allegiance to ANSI/ASQ, and appropriate risk-consideration), might be to, divide the 80000pcs into separate "receiving batches" (perhaps 3 sets of 20000), and perform the same exercise on each.
     
    Vthouta likes this.
  3. Mike S

    Mike S Member

    Joined:
    Jan 11, 2016
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Thank you, Mark. Are you saying that in ANSI Z1.9 (or MIL-STD-414) there's no typical carve-out or procedure for re-submitting or re-sampling the non conforming lot?
     
  4. Vthouta

    Vthouta Member

    Joined:
    Dec 20, 2015
    Messages:
    10
    Likes Received:
    2
    Trophy Points:
    2
    Vaguely, I remember few years back, I faced the similar scenario on one of product. We had only one defect for AQL level 2.0, but then I used tightened inspection to confirm if the team had any similar defects. This helped us to estimate and give us more brief idea on to accept the lot or not. But, at the end of the day, as Mark said it depends on the risk and urgency to get out the good product of the door.
     
  5. Miner

    Miner Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    258
    Likes Received:
    186
    Trophy Points:
    42
    Location:
    Greater Milwaukee USA
    Separate from the specific question you asked regarding Z1.9, I would like to address the outlier itself. Was this just the tail area of the distribution, or was it a true outlier that was significantly separated from the other 149 samples? If it was a true outlier, did you recheck the measurement to rule out a measurement anomaly, transposed digits or recording error?
     
  6. Mike S

    Mike S Member

    Joined:
    Jan 11, 2016
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Mean = 90.36, STDEV = 2.350. The outlier was 105.66. That's over 6.5 Sigma to one side. Next highest value was 93.23. Upper limit is 95.25. Sure looks like an outlier to me.

    We rechecked the measurement, and it's legit. This same sample also failed the upper limit on a separate metric, but didn't result in a non-compliance on that metric.
     
  7. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    362
    Likes Received:
    359
    Trophy Points:
    62
    Location:
    Maine
    Ok if the measurement is good and the test was valid it wasn't an outlier in the sense that you can delete it from the data set. It was a real value. I assume that the inspection failed because the SD indicates that there are several defects in the lot? In this case the continuous sampling breaks down because you probably do not have a Normal distribution. The two approaches *I* use are to revert to categorical sampling which is not distributional dependent or investigate splitting the inspection lot up to determine which part has the (potentially) high defect rate.
     
  8. Miner

    Miner Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    258
    Likes Received:
    186
    Trophy Points:
    42
    Location:
    Greater Milwaukee USA
    Not knowing what the measured characteristic is, or how it is manufactured, we can only speculate, but it is possible that this was caused by setup parts or mixed process streams. I would go into investigative mode, take a larger sample and plot the results rather than try different sampling plans. If you accept the lot and there are more units like this, what will be the impact to your process and your customer?
     
    Bev D likes this.
  9. QAengineer13

    QAengineer13 Member

    Joined:
    Aug 3, 2015
    Messages:
    47
    Likes Received:
    14
    Trophy Points:
    7
    IMHO , in your investigation mode also identify if possible from Batch records is there any change and perform a RCA to rule out that its not Process /Rawmaterial/Equiment/Measurement/Personnel related!
     
  10. DavidD

    DavidD Member

    Joined:
    Sep 30, 2015
    Messages:
    24
    Likes Received:
    17
    Trophy Points:
    2
    Within MIL-STD-414, there were provisions allowing for utilizing attribute sampling (MIL-STD-105) when the sample failed the variable acceptance. I don't remember all the details but from a quick look it seems to be arround sections A9.3 and A9.4 (page #3). I don't know if there are similar provisions within Z1.9 (I don't have a copy at the moment).

    As an aside, more modern plans (MIL-STD-1916 or ISO 21247) don't allow for "reresting/reverting" to attribute when you fail variable.

    David
     
  11. Mike S

    Mike S Member

    Joined:
    Jan 11, 2016
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Very interesting DavidD. Here's what MIL-STD-414 says in Section A9.2.2:

    A9.2.2 Mixed Variables-Attributes Inspection
    Mixed variables-attributes inspection is inspection of a sample by attributes, in addition to inspection by variables already made of a previous sample, before a decision as to acceptability or rejectability of a lot can be made.

    Here's what Z1.9 says in A2.5:

    A2.5 Mixed Variables-Attributes Inspection: Mixed variables-attributes inspection is inspection of a sample by attributes, in addition to inspection by variables already made of a previous sample from the lot, before a decision as to acceptability or rejectability of a lot can be made.

    So there is a carve-out in MIL-STD-414 and Z.19 to convert to a MIL-STD-105E or Z1.4 plan (with an increase in sample size, obviously) as a "re-submission" of the lot.
     
  12. Mike S

    Mike S Member

    Joined:
    Jan 11, 2016
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Another question: If I am switching to Tightened Inspection, how do I execute that? I couldn't find an example of tightened inspection in Z1.9. I am using Table B-3 (Double Specification Limit, Variability Unknown. Previously, under Normal Inspection, I was Sample Code N, AQL = 0.65, M = 1.42. What changes when I go to Tightened?
     
  13. DavidD

    DavidD Member

    Joined:
    Sep 30, 2015
    Messages:
    24
    Likes Received:
    17
    Trophy Points:
    2
    I don't recall having alot of personal experience with switching w/ Z1.9 or 414 but taking a quick look, it appears that 414 uses a separate AQL scale at the bottom of the charts for tightened. In effect, what would have been done with an 0.65 AQL for normal now shifts a column to the left and uses the 0.40 AQL requirements (I'm looking at table B-3 on page 45 of MIL-STD-414).

    David
     
  14. MikeH

    MikeH Member

    Joined:
    Feb 23, 2016
    Messages:
    7
    Likes Received:
    3
    Trophy Points:
    2

Share This Page