Determining the number of operational modes in baseline multivariate SPC data

Hang Zhang, Susan Albin

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

When building a multivariate SPC model, it is commonly assumed that there is only one operational mode in the baseline data. However, multiple operational modes may exist. It is important to know the number of modes in the data in order to construct an effective process control system. Each operational mode appears as a cluster in the baseline data. This paper proposes a method to identify the correct number of clusters in baseline data. None of the existing methods for finding the number of clusters has all three of the following critical features: (i) the proposed method can determine if there is only one cluster, the most common case in baseline data; (ii) it can identify clusters that are convex or non-convex; and (iii) it is insensitive to user-specified parameters. The paper includes a comparison of the existing and proposed methods on four datasets. The proposed method consistently gives the correct number of clusters whereas the existing methods are unable to do so.

Original languageEnglish (US)
Pages (from-to)1103-1110
Number of pages8
JournalIIE Transactions (Institute of Industrial Engineers)
Volume39
Issue number12
DOIs
StatePublished - Dec 1 2007

Fingerprint

Process control
Control systems

All Science Journal Classification (ASJC) codes

  • Industrial and Manufacturing Engineering

Keywords

  • Clustering
  • Data mining
  • Multivariate statistical process control
  • Number of clusters
  • Phase I

Cite this

@article{e3b678448c724ddcbbcae5939f487081,
title = "Determining the number of operational modes in baseline multivariate SPC data",
abstract = "When building a multivariate SPC model, it is commonly assumed that there is only one operational mode in the baseline data. However, multiple operational modes may exist. It is important to know the number of modes in the data in order to construct an effective process control system. Each operational mode appears as a cluster in the baseline data. This paper proposes a method to identify the correct number of clusters in baseline data. None of the existing methods for finding the number of clusters has all three of the following critical features: (i) the proposed method can determine if there is only one cluster, the most common case in baseline data; (ii) it can identify clusters that are convex or non-convex; and (iii) it is insensitive to user-specified parameters. The paper includes a comparison of the existing and proposed methods on four datasets. The proposed method consistently gives the correct number of clusters whereas the existing methods are unable to do so.",
keywords = "Clustering, Data mining, Multivariate statistical process control, Number of clusters, Phase I",
author = "Hang Zhang and Susan Albin",
year = "2007",
month = "12",
day = "1",
doi = "10.1080/07408170701291787",
language = "English (US)",
volume = "39",
pages = "1103--1110",
journal = "IISE Transactions",
issn = "2472-5854",
publisher = "Taylor and Francis Ltd.",
number = "12",

}

Determining the number of operational modes in baseline multivariate SPC data. / Zhang, Hang; Albin, Susan.

In: IIE Transactions (Institute of Industrial Engineers), Vol. 39, No. 12, 01.12.2007, p. 1103-1110.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Determining the number of operational modes in baseline multivariate SPC data

AU - Zhang, Hang

AU - Albin, Susan

PY - 2007/12/1

Y1 - 2007/12/1

N2 - When building a multivariate SPC model, it is commonly assumed that there is only one operational mode in the baseline data. However, multiple operational modes may exist. It is important to know the number of modes in the data in order to construct an effective process control system. Each operational mode appears as a cluster in the baseline data. This paper proposes a method to identify the correct number of clusters in baseline data. None of the existing methods for finding the number of clusters has all three of the following critical features: (i) the proposed method can determine if there is only one cluster, the most common case in baseline data; (ii) it can identify clusters that are convex or non-convex; and (iii) it is insensitive to user-specified parameters. The paper includes a comparison of the existing and proposed methods on four datasets. The proposed method consistently gives the correct number of clusters whereas the existing methods are unable to do so.

AB - When building a multivariate SPC model, it is commonly assumed that there is only one operational mode in the baseline data. However, multiple operational modes may exist. It is important to know the number of modes in the data in order to construct an effective process control system. Each operational mode appears as a cluster in the baseline data. This paper proposes a method to identify the correct number of clusters in baseline data. None of the existing methods for finding the number of clusters has all three of the following critical features: (i) the proposed method can determine if there is only one cluster, the most common case in baseline data; (ii) it can identify clusters that are convex or non-convex; and (iii) it is insensitive to user-specified parameters. The paper includes a comparison of the existing and proposed methods on four datasets. The proposed method consistently gives the correct number of clusters whereas the existing methods are unable to do so.

KW - Clustering

KW - Data mining

KW - Multivariate statistical process control

KW - Number of clusters

KW - Phase I

UR - http://www.scopus.com/inward/record.url?scp=36148930994&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=36148930994&partnerID=8YFLogxK

U2 - 10.1080/07408170701291787

DO - 10.1080/07408170701291787

M3 - Article

AN - SCOPUS:36148930994

VL - 39

SP - 1103

EP - 1110

JO - IISE Transactions

JF - IISE Transactions

SN - 2472-5854

IS - 12

ER -