Small Sample Performance of Some Statistical Setup Adjustment Methods

Abstract

The setup adjustment problem occurs when a machine experiences an upset at setup that needs to be compensated for. In this article, feedback methods for the setup adjustment problem are studied from a small-sample point of view, relevant in modern manufacturing. Sequential adjustment rules due to Grubbs (Grubbs, F. E. (1954). An optimum procedure for setting machines or adjusting processes. Industrial Quality Control 07) and an integral controller are considered. The performance criteria is the quadratic off-target cost incurred over a small number of parts produced. Analytical formulae are presented and numerically illustrated. Two cases are considered, the first one where the setup error is a constant but unknown offset and the second one where the setup error is a random variable with unknown first two moments. These cases are studied under the assumption that no further shifts occur after setup. It is shown how Grubbs' harmonic rule and a simple integral controller provide a robust adjustment strategy in a variety of circumstances. As a by-product, the formulae presented in this article allow to compute the expected off-target quadratic cost when a sudden shift occurs during production (not necessarily at setup) and the adjustment scheme compensates immediately after its occurrence.

Keywords:

• Process adjustment
• Stochastic approximation
• Integral control

Acknowledgment

This article was partially funded by NSF grant DMI 9996031.