The statistical quality control of a process consists of two phases. In phase I, preliminary data of 20–30 samples of about 5 units each are collected to construct control charts, such as and R charts. The limits computed from these preliminary samples are referred to as trial control limits and they enable us to determine whether the process is in-control when the initial samples are selected. These control charts are then used to monitor for problems in a future process (i.e., phase II monitoring) such as outliers or excess variability in the subgroups which may indicate the presence of an assignable cause. The presence of outliers in the preliminary samples will stretch the trial limits of conventional charts such as and R charts, making them difficult in detecting these outliers. As a result, a bulk of these outliers will be included in the estimation of the trial limits so that the computed limits become wider and less meaningful. Thus, the detection of any out-of-control (o.o.c.) behaviour in phase II becomes slower and more difficult. In this paper, two time weighted robust control charts for the process variance which allow quicker detection of outliers and increased sensitivity to other types of o.o.c. signals when outliers are present are presented. These new approaches use a weighted average and an unweighted moving average as the charts' statistics. A robust estimator of scale, i.e., the sample interquartile range is used in the computation of the control limits. It is shown that overall better performances are provided by these two charts in comparison to that of the existing robust chart for process variance which uses the same estimator of scale.
[1]
Lawrence G. Tatum.
Robust estimation of the process standard deviation for control charts
,
1997
.
[2]
Kit C. B. Roes,et al.
Robust Individuals Control Chart for Exploratory Analysis
,
2004
.
[3]
Moustafa Omar Ahmed Abu-Shawiesh,et al.
NEW ROBUST STATISTICAL PROCESS CONTROL CHART FOR LOCATION
,
1999
.
[4]
Moustafa Omar Ahmed Abu-Shawiesh,et al.
ESTIMATING THE PROCESS STANDARD DEVIATION BASED ON DOWNTON'S ESTIMATOR
,
2000
.
[5]
David M. Rocke.
Robust control charts
,
1989
.
[6]
M.B.C. Khoo.
Some control charts for the process mean and variance based on Downton's estimator
,
2004,
2004 IEEE International Engineering Management Conference (IEEE Cat. No.04CH37574).
[7]
Benjamin M. Adams,et al.
Robust Monitoring of Contaminated Data
,
2005
.
[8]
David M. Rocke.
Xq and Rq charts: Robust control charts
,
1992
.
[9]
Maria E. Calzada,et al.
THE ROBUSTNESS OF THE SYNTHETIC CONTROL CHART TO NON-NORMALITY
,
2001
.
[10]
Harrison M. Wadsworth,et al.
Modern methods for quality control and improvement
,
1986
.
[11]
Benjamin M. Adams,et al.
Alternative Designs of the Hodges-Lehmann Control Chart
,
1996
.
[12]
J. Lucas,et al.
Robust cusum: a robustness study for cusum quality control schemes
,
1982
.
[13]
Mukund Raghavachari,et al.
Control chart based on the Hodges-Lehmann estimator
,
1991
.