[ | Next | Previous | Up ]

Re: sample plan for spindle discs

From: Stan Hilliard
Date: 19 May 2000
Time: 17:39:02

Comments

Greetings Dave,

I see two issues here. 1) how to efficiently determine if a spindle contains 150 disks, and 2) how many spindles to sample per lot.

If the variability of spindle tare weights is less than the weight of one disk, you could weigh full spindles to determine the count. You can determine if this criterion is met with an Xbar and R chart for tare weights over a series of lots. If all goes well, you can set acceptance limits for the weight of individual spindles with a sampling plan for the mean.

Use AQL (mean)=target weight for tare + 150 disks.

Use the lower RQL (mean)=AQL (mean) - the weight of one disk.

Use the upper RQL (mean)=AQL (mean) + the weight of one disk.

See the example here:

www.samplingplans.com/outputvariablesmean.htm

For a more thorough tutorial:

www.samplingplans.com/modern3.htm

The sampling plan for the mean will require known standard deviation, measured as the standard deviation of good (only) spindles containing 150 disks.

The sampling plan will contain the decision limits for verifying a single spindle. If the calculated n=1 then you have a procedure that can detect individual miscounts. You could apply this method to each spindle before it is used in process.

If you want to use lot sampling to avoid testing all the spindles, I think attribute sampling is appropriate because the defectives are not from the same distribution of good spindles as in the plan for the mean. Using TP105, set AQL and RQL fraction defective to control the protection.

Sincerely, Stan Hilliard


Last changed: November 20, 2007