Recognition of unnatural patterns in manufacturing processes using the Minimum Description Length criterion

Research output: Contribution to journalArticle

Abstract

This paper proposes a method for monitoring processes, based on the Minimum Description Length (MDL) criterion, which can detect unnatural events, recognize unnatural patterns and estimate the occurring point automatically and simultaneously. Several studies have investigated the application of the MDL criterion to statistical process control. This paper extends this investigation to the problem of recognizing unnatural patterns on process control. First, the equations of the MDL criterion corresponding to the underlying unnatural patterns (shifts, trends and peaks) are derived. Next, an algorithm for monitoring processes based on the derived MDL criterion is provided. Finally, simulation experiments are conducted to examine the validity of the proposed method under various conditions.

Original languageEnglish
Pages (from-to)583-601
Number of pages19
JournalCommunications in Statistics - Theory and Methods
Volume29
Issue number2
Publication statusPublished - 1999 Dec 1

    Fingerprint

Keywords

  • Information theory
  • Model selection
  • Statistical process control

ASJC Scopus subject areas

  • Statistics and Probability

Cite this