A compressed sensing feature extraction approach for diagnostics and prognostics in electromagnetic solenoids

  • One major realm of Condition Based Maintenance is finding features that reflect the current health state of the asset or component under observation. Most of the existing approaches are accompanied with high computational costs during the different feature processing phases making them infeasible in a real-world scenario. In this paper a feature generation method is evaluated compensating for two problems: (1) storing and handling large amounts of data and (2) computational complexity. Both aforementioned problems are existent e.g. when electromagnetic solenoids are artificially aged and health indicators have to be extracted or when multiple identical solenoids have to be monitored. To overcome those problems, Compressed Sensing (CS), a new research field that keeps constantly emerging into new applications, is employed. CS is a data compression technique allowing original signal reconstruction with far fewer samples than Shannon-Nyquist dictates, when some criteria are met. By applying this method to measured solenoid coil current, raw data vectors can be reduced to a way smaller set of samples that yet contain enough information for proper reconstruction. The obtained CS vector is also assumed to contain enough relevant information about solenoid degradation and faults, allowing CS samples to be used as input to fault detection or remaining useful life estimation routines. The paper gives some results demonstrating compression and reconstruction of coil current measurements and outlines the application of CS samples as condition monitoring data by determining deterioration and fault related features. Nevertheless, some unresolved issues regarding information loss during the compression stage, the design of the compression method itself and its influence on diagnostic/prognostic methods exist.

Export metadata

Additional Services

Share in Twitter Search Google Scholar
Metadaten
Author:Christian Knöbel, Hanna Wenzl, Johannes ReuterORCiD, Clemens GühmannGND
URL:http://www.phmsociety.org/events/conference/phm/17/proceedings
ISBN:978-1-936263-26-4
Parent Title (English):PHM 2017 : proceedings of the Annual Conference of the Prognostics and Health Management Society 2017 : St. Petersburg, Florida, USA, October 2-5, 2017
Publisher:PHM Society
Document Type:Conference Proceeding
Language:English
Year of Publication:2017
Release Date:2019/08/05
First Page:147
Last Page:152
Open Access?:Ja
Nein
Relevance:Keine peer reviewed Publikation (Wissenschaftlicher Artikel und Aufsatz, Proceeding, Artikel in Tagungsband)