Robust Monitoring of Contaminated Data

Abstract

Monitoring a process that has contaminated data with traditional control charts such as Shewhart's X̄ chart and the Range chart results in an excessive number of false alarms. Robust control charts such as the Median and IQR charts are a better alternative to traditional charts for a process with contaminated data because the effects of the outlying data values are eliminated. However, process shifts are not detected as quickly with the robust charts. This paper introduces a diagnostic statistic technique that uses traditional control chart methods augmented by diagnostic tools. Performance measures for traditional, robust, and diagnostic statistic control chart systems for n = 5 are reported. Contour plots for n = 3 and n = 5 are provided to allow for interpolation of parameter values. The diagnostic statistic technique improves work stoppage (comparable to average run length) rates for contaminated data and maintains the ability to identify process shifts.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 20 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.