Skip to main navigation Skip to search Skip to main content

Kalman filter based neural network methodology for predictive maintenance: A case study on steam turbine blade performance prognostics

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

Abstract

Predictive maintenance involves condition monitoring, fault detection and prediction of remaining useful life or forthcoming failures. Predictive maintenance systems for steam turbine engines offer detection, classification, and prediction (or prognosis) of potential critical component failures, and ensures substantially reducing the cost of repair and replacement of defective parts, and may even result in saving lives. This paper describes a Kalman filter based neural network approach to provide performance evaluation and residual life prediction with the objectives of improving availability and implementing maintenance before failure occurs by estimating degradation severity and the proper timing for replacement. The approach has been applied to a steam turbine blade fatigue experiment testbed to illustrate the prognostic functionalities of the methodology.

Original languageEnglish
Title of host publicationProceedings of 2006 ASME International Mechanical Engineering Congress and Exposition, IMECE2006 - Manufacturing
PublisherAmerican Society of Mechanical Engineers (ASME)
ISBN (Print)0791837904, 9780791837900
DOIs
StatePublished - 2006
Event2006 ASME International Mechanical Engineering Congress and Exposition, IMECE2006 - Chicago, IL, United States
Duration: 5 Nov 200610 Nov 2006

Publication series

NameAmerican Society of Mechanical Engineers, Manufacturing Engineering Division, MED
ISSN (Print)1071-6947

Conference

Conference2006 ASME International Mechanical Engineering Congress and Exposition, IMECE2006
Country/TerritoryUnited States
CityChicago, IL
Period5/11/0610/11/06

Keywords

  • Fatigue damage
  • Kalman filter
  • Neural network
  • Prognostics
  • Steam turbine

Fingerprint

Dive into the research topics of 'Kalman filter based neural network methodology for predictive maintenance: A case study on steam turbine blade performance prognostics'. Together they form a unique fingerprint.

Cite this