Complex sample drawing

Logistics QM

Moderators: Snowy, thx4allthefish, Ha Tran

Post Reply
Posts: 970
Joined: Mon Oct 21, 2002 1:53 am
Location: Norway

Complex sample drawing

Post by Trygve » Tue Jul 08, 2014 4:31 am


I've got an issue with a sample drawing. It's for some agricultural product.

At GR a number of 20 kg samples are drawn based on the size of the delivery. That can be fixed using sample drawing procedure and sampling scheme.
Then the samples are sorted and those too small or large removed, thus the sample is then normally around 19 kg.

Then 10 kg are taken out from each sample for testing. Based on each sample an 'Error score' is calculated. So far so good.

The challenge comes if one or more of the samples has an error score in the interval that makes a retest required.
The remainder of the sample, aproximately 9 kg is then tested.

My challenges are :
- Customer want's this additional result recording appear automatically
- The value of the characteristics should be calculated as an average of the samples where the values for thesample(s) that have been retested is a weighted average to the two tests.

Thus if everything is OK then the value for the batch is (A + B + C ) / 3 and if retest of A : [(A1*10/10 + A2*10/x)/2 + B + C ] / 3 where x is the weight of A2.

As I see it I can, in the insp.lot create (manually) another phys.sample for the insp lot to cover the retest, but the system will then concider it a separate asmple and thus calculate (A1 + A2 + B + C) / 4

I can add all the MIC's twice in the insp.plan at the bottom and try setting them 'After rejection' based on the 'Error score' but the system then would not know how to handle the double results and since I'm using referenced MIC's to get the result transferred to the batch I'll have to unlock to be able to change the control indicators for the MIC.

I can create additional MICs just for the retest, but then the system wouldn't know that it's the same analysis, and thus not calculate correct.

I'm running out of ideas. Any suggestions on how to handle this issue would be highly appeciated.

Post Reply