Skip to main navigation Skip to search Skip to main content

Study on KPI-related subspace decomposition for fault detection and robust KPI prediction against abnormal data

  • Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper deals with key performance indicator (KPI) related fault detection system design issues based on an integrated framework. The characteristics of the projection directions and the extracted subspaces are discussed in details. In comparison with the basic PLS based approach, the proposed algorithm improves the decomposition performance with respect to KPI. In the meantime, by incorporating the partial robust M-regression (PRM) algorithm into the expectation maximum (EM) framework, abnormal data including outliers and missing data are carefully dealt with. At last, a demonstration on the industrial benchmark, the Tennessee Eastman process (TEP) is provided to verify the validity and superiority of the proposed method.

Original languageEnglish
Title of host publicationProceedings - 2016 IEEE 25th International Symposium on Industrial Electronics, ISIE 2016
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages100-105
Number of pages6
ISBN (Electronic)9781509008735
DOIs
StatePublished - 17 Nov 2016
Event25th IEEE International Symposium on Industrial Electronics, ISIE 2016 - Santa Clara, United States
Duration: 8 Jun 201610 Jun 2016

Publication series

NameIEEE International Symposium on Industrial Electronics
Volume2016-November

Conference

Conference25th IEEE International Symposium on Industrial Electronics, ISIE 2016
Country/TerritoryUnited States
CitySanta Clara
Period8/06/1610/06/16

Fingerprint

Dive into the research topics of 'Study on KPI-related subspace decomposition for fault detection and robust KPI prediction against abnormal data'. Together they form a unique fingerprint.

Cite this