Identifying and visualizing nonlinear variation patterns in multivariate manufacturing data

Daniel W. Apley*, Feng Zhang

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

7 Scopus citations


In modern manufacturing processes, it is not uncommon to have hundreds, or even thousands, of different process variables that are measured and recorded in databases. The large quantities of multivariate measurement data usually contain valuable information on the major sources of variation that contribute to the final product or process variability. The focus of this work is on variation sources that result in complex, nonlinear variation patterns in the data. We propose a model for representing nonlinear variation patterns and a method for blindly identifying the patterns, based on a sample of measurement data, with no prior knowledge of the nature of the patterns. The identification method is based on principal curve estimation, in conjunction with a data preprocessing step that makes it suitable for high dimensional data. We also discuss an approach for interactively visualizing the nature of the identified variation patterns, to aid in identifying and eliminating the major root causes of manufacturing variation.

Original languageEnglish (US)
Pages (from-to)691-701
Number of pages11
JournalIIE Transactions (Institute of Industrial Engineers)
Issue number6
StatePublished - Jun 2007


  • Blind identification
  • Principal component analysis
  • Principal curves
  • Statistical process control
  • Variation patterns

ASJC Scopus subject areas

  • Industrial and Manufacturing Engineering


Dive into the research topics of 'Identifying and visualizing nonlinear variation patterns in multivariate manufacturing data'. Together they form a unique fingerprint.

Cite this