Vinod GC

In a cold rolling mill, a sample is taken for inspection at a frequency of one in 10 coils. This frequency was not arrived at scientifically (i.e without any logic). On an average 152 coils are rolled in a 24 hours shift. This sampling frequency eats up on the production capacity. On a typical day the rolling schedule contains a wide variety of SKU's (alloy, width, gauge, temper etc) depending on the delivery priorities. Historical data pertaining to the defects identified on the samples are available. How do we arrive at an optimum sampling frequency based on historical data?