Top Page | Upper Page | Contents | About This Site | JAPANESE

Sampling Inspection

In factories, there are inspections at the middle and final process.

If the unit is the lot, the inspection should be done for all products. But sampling is used in many cases because of two reasons.

The first reason is that products are broken by the destructive testing.

The second reason is economic reason. If the inspection needs long time, it delays the delivery. Or if the inspection needs much money, the more inspection, the higher the cost.

Below is the some methods related to Process Analysis for Normal Condition .

And below is the methods for the sampling more than two samples.

Prediction of All Products by Sampling Data

Prediction interval is useful to predict the value of possibility of NG products.

Prediction of the Yield

The data of all products are needed to calculate the exacct value of the yield.

But we can predict the value by the application of the logic of the prediction interval .

Calculation of the Predicted Yield

• 1. Average of samples(Ave), standard deviation(Std), number of samples(n), higher range of Go(High), lower range of Go(Low)
• 2. t-value of "High" = (High - Ave) / Std / SQRT(1 + 1/n)
  t-value of "Low" = (Ave - Low) / Std / SQRT(1 + 1/n)
  SQRT() means the square root.
• 2. If using EXCEL,
  Proportion defective of "High" = TDIST(t-value of "High" ,n-1,1)
  Proportion defective of "Low" = TDIST(t-value of "Low" ,n-1,1)
• 3. Yild = 1 - Proportion defective of "High" - Proportion defective of "Low"

NEXT Process Analysis for Abnormal Condition