00230 819cd9235a78d0f202ffb72a6ad790

232

Baxley

Figurę 10. Fit of IMA Model to Machinę Average Y₂ Data from June 1988.

Process engineers can access the system in order to tune the EWMA controller by changing L, X, g, or a_A. Historical data files of (1) tuning adjustments, (2) Shewhart rule violations, and (3) Yj and pressure variation are being maintained; and a menu driven system for analysis of the latter file has been written using SAS/AF® and SAS/QC® software.

The main menu for this monitoring program is shown in Figurę 11, and it can be used to evaluate the performance of the EWMA controller. Soon after the system was started it became apparent that, as shown by simulation studies [Baxley(1990)J and in the controller tuning section of this paper, setting L = 0 gave the least variability in Y^ This setting was then implemented because, unlike the manuał system, there is no cost associated with making the pressure adjustments morę freąuently. Values of X from .3 to .5 were tested on different machines, and after an additional month it became apparent that .3 was a good conservative setting for avoidance of over-adjustment. Electing menu option #4 gives the run chart of ^ and pressure variability in Figurę 12. This chart is scaled based on knowledge of the process gain, and it can be seen that almost alf of the process drifts are being absorbed by the controller adjustments, leaving mostly common-cause variation in the Yj data.