STATISTICAL MEASUREMENT DEVIATIONS IN QUALITY-CONTROL SYSTEMS

The selectivity enabling quality control systems to recognize good and bad parts, is seriously affected by its measurement deviations. This observation holds even with no systematic but only statistical measurement deviations present. Discussion is based on known statistical models and distribution of production data. Four different selection criteria are used: I) correct selection of bad parts, II) incorrect selection of bad parts, III) correct selection of good parts and IV) incorrect selection of good parts. Numerical integration is applied in order to calculate these four different contributions as a function of the difference between the selection threshold and production mean values, considering standard deviations of measurement as well as production process as parameters. It will be shown that the number of good products incorrectly selected as bad is larger than the number of correctly selected bad products if on one hand the measurement standard deviation is almost as large as the standard deviation of production and on the other hand the threshold is close to the production mean value.