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.

Minimum receiving inspection

Discussion in 'Sampling, Standards and Inspection' started by APanteva, Sep 14, 2019.

  1. APanteva

    APanteva Member

    Joined:
    Jul 13, 2019
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    I am looking for the most optimal receiving inspection plan, because the test is time consuming.
    Our vendor claims their outgoing reject rate is 0.03%. We receive batches of 900 pcs at a time.
    If I select AQL=0.25 (the defect is critical), per double sampling plan the sampling size in best case should be 50/100. Is there a way for me to lower the sampling size without increasing the AQL too much?
     
  2. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    458
    Likes Received:
    494
    Trophy Points:
    62
    Location:
    Maine
    When you have a critical defect and need a low defect rate - or no defects - inspection is not useful. It will only protect you if the defect rate spikes in a lot.
    You would benefit more by focusing on controls at the supplier for the process inputs and SPC on continuous data if possible or automated visual inspection of some type if the defect is truly categorical in nature.

    Also remember that the AQL is ACCEPTABLE defect rate. If this is a critical defect you are probably better off with an RQL plan bu that would increase your sample size even more so you are back to my first recommendation.
     
  3. APanteva

    APanteva Member

    Joined:
    Jul 13, 2019
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Thank you, Bev D. Appreciated! I knew it was a long shot. The manufacturer has 100% test already and says the end defect rate is 0.03%, but after testing about 200 items we found the defect ate to be 1.5%. I just want to make sure that a lot with defect rate higher than 0.25 is rejected. Thank you.
     
  4. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    458
    Likes Received:
    494
    Trophy Points:
    62
    Location:
    Maine
    25% (.25) or 0.25% or?

    the need to reject lots at .25% or higher requires a sample size of 1565 c=0
     
  5. APanteva

    APanteva Member

    Joined:
    Jul 13, 2019
    Messages:
    5
    Likes Received:
    0
    Trophy Points:
    1
    Thank you for taking the time, Bev. 0.25% is correct. We sell in big volumes from the manufacturer straight to the customer. I am confused, though. I was looking into single sampling plan normal inspection II (Z1.4) and for 900 lot size (J) the sample 80 pcs C=0 for AQL=0.25. Is that correct?
     
  6. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    458
    Likes Received:
    494
    Trophy Points:
    62
    Location:
    Maine
    I'm sure that is the correct sample size for an AQL of .25 and a lot size of 900.

    However, you stated that the rate you wanted to be sure to reject was .25%. That requires an RQL plan. Remember that AQL stands for ACCEPTABLE quality level. In other words you will ACCPET a lot with a .25% defect rate 95% of the time.
     
  7. Joanne1

    Joanne1 New Member

    Joined:
    Oct 16, 2019
    Messages:
    2
    Likes Received:
    0
    Trophy Points:
    1
    Hi Bev D. I like your Inspection Sampling Plan for Categorical Data - RQL and AQL Based Plans Excel Spreadsheets. I update our and generate a sampling plan. I don't know how it was determined our current sampling plan. It is an AQL 2.5, ANSI Z1.4,2003, Table II, but it is not falling into the Table II, A, B, C. How can I determine a new sampling plan? What is the meaning of P=0.02 and Confidence =0.98 , and n= 194 in your Binomial Equation. I am seeing also a Confidence table with P%, p, 95%, 99%, ppm. And an n= 196 with p, P(Miss), P(Detect), and Sample Size with Max p, 95%, 99%. I need to create a matrix and a sampling size system and I am not sure if I should clarify how to implement our current AQL 2.5 Table II ANSI Z1.4 or a better sampling size table.

    Our sampling size are:

    AQL 2.5 Table II ANSI

    Lot >=2-50, inspect 5 parts
    Lot >= 51-150, inspect 20 parts
    Lot >=151-280, inspect 32 parts
    Lot >=281-500, inspect 50 parts
     
  8. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    458
    Likes Received:
    494
    Trophy Points:
    62
    Location:
    Maine
    these are just example values, you can put in any values you want for p and Confidence. p= the defect rate as a proportion (not a percentage) and confidence is the confidence level you want to DETECT the defect rate p. so in this example you need a sample size of n=194 to detect a defect rate of .02 (2%) 98% of the time. (the n=194 is form the Binomial equation while the Poisson gives a sample size of 196. the two equations yield approximately the same sample size and will converge tot he same as the defect rate approaches 0. So to determine a new sample size just enter the defect rate you want to detect (this is the RQL also known as the LTPD not the AQL) and the confidence you want (also known as the probability of detection, P(D)).

    The table on the left shows the sample sizes for various defect rates (expressed as a proportion (p) and a percentage P% and a ppm. This table uses Poisson equation. (note there is a typo in this table that I just noticed. the P% is not 5%, it is 10%)

    The middle table is a tabular version of the OC curve for the Poisson generated sample size of 196 for a defect rate of .02 and 98% confidence. it just shows the probability of detecting and missing the given defect rates. This table will change as you enter your own p and confidence numbers.

    The table on the right "Max p" is just another look at sample sizes needed for various defect rates so you can compare various plans.

    You can use this tab to create your sample tables. Note that lot size has no bearing on the sample size (unless your lot is really small and then you need to use the hypergeometric, not the poisson or binomial) even though many popular tables include them. They are a compromise from the OLD days (1950s). also note that this tab uses the RQL or the defect rate that you want to detect/reject. the tables you refer to use the AQL which is the defect rate you want to ACCEPT most of the time. A further note is that this tab is a c=0 plan.

    IF you need AQL and RQL indexed plans then use the AQL/RQL tab.
     

Share This Page