00232 Þ9b1ac6d3646bcdf4e254edd2dae625

234

Baxley

Choosing menu option #5 results in an individual and moving rangÄ™ of 2 (IR) chart of the forecast errors as shown in Figurel3. In the case where L- 0 an adjustment is madÄ™ for every sample, and the forecast errors (A_t) are equal to the control errors (Y₍ - Target). The control errors are seen to be in statistical control except for a run of seven below target which coincides with a relatively fast increase in pressure. This was caused by a sudden shift in the level of Yj, which is a departure from the norma! drifting behavior, and it highlights the utility of the modified Shewhart chart with Ford rules for detecting special causes in a system under EWMA control. Another possibility for automated systems such as this is use of the SPC methods of Alwan and Roberts (1988) on the highly auto-correlated adjustment data.

For menu option #6, the user enters a value for process gain, and then a process capability study is performed on all machines, giving results as shown in Table 1. The open-loop sigma was determined by backing the eflfect of the pressure adjustments out of the Y₂ data based on the process gain just entered.

CONTROL CHARTS OF A₂ AND MOVING RANGĘ MACH INE = P 1 4

3<*L iłii t s For o=2:

Subęroup lndox (DATETIME)

FigurÄ™ 13. Display for Menu Choice S of Y₂ Monitoring Program.