00213 ö63458cc28250959e6eae16310a0980

11

Applications of the EWMA for Algorithmic Statistical Process ControP

Robert V. Baxley, Jr. Monsanto Chemical Co. Pensacola, Florida

Abstract

Use of the ExponentialIy Weighted Moving Average (EWMA) statistic is considered for implementation of a statistical control strategy for drifting processes. This strategy keeps the process centered at the target by making process adjustments while at the same time it detects sudden large shifts for which troubleshooting is necessary. In the past, Monsanto has typically used a modified Shewhart chart with runs rules to determine process adjustments. This new EWMA strategy is shown to be an improvement because of better tuning capability, robustness to incorrect tuning, and a decreased tendency to cause additional variability with large, inappropriate adjustments. These improvements are illustrated in two case studies involving important nylon carpet fiber quality parameters.

*This chapter is reprinted, with permission, from Quaiity Engineering, 1994-1995, vol. 7, no. 2, pp. 397-418.

215