The decay of a single component in a naval propulsion system can affect the performance of the entire system, involving expensive maintenance costs for restoring efficient conditions. Therefore, a regular control of the decay of key components of these systems is essential for properly handle maintenance actions. Moreover, in naval propulsion systems it is necessary to consider the difficulty in implementing an onboard maintenance action or returning a vessel. Two relevant components in naval propulsion systems are the turbine and the compressor. This study develops two machine learning models to predict turbine and compressor decay, i.e. based on classification and regression approaches. The former classifies whether the components are decayed or not, thus highlighting a state of criticality, the latter predicts a specific value of each decay coefficient. For each approach, different algorithms are compared, e.g. boosted trees, linear regression or support vector machines. A case study considering sixteen inputs has been used to test the effectiveness of the proposed solution, starting from a dataset of about twelve thousand instances referred to a naval vessel. A sensitivity analysis of relevant parameters has been developed to verify the robustness of the approach.

Machine learning models to predict components decay in a naval propulsion system / Quatrini, E.; Colabianchi, S.; Costantino, F.; Tronci, M.. - In: ...SUMMER SCHOOL FRANCESCO TURCO. PROCEEDINGS. - ISSN 2283-8996. - (2020). (Intervento presentato al convegno 25th Summer School Francesco Turco tenutosi a Bergamo, Italy).

Machine learning models to predict components decay in a naval propulsion system

Quatrini E.
;
Colabianchi S.;Costantino F.;Tronci M.
2020

Abstract

The decay of a single component in a naval propulsion system can affect the performance of the entire system, involving expensive maintenance costs for restoring efficient conditions. Therefore, a regular control of the decay of key components of these systems is essential for properly handle maintenance actions. Moreover, in naval propulsion systems it is necessary to consider the difficulty in implementing an onboard maintenance action or returning a vessel. Two relevant components in naval propulsion systems are the turbine and the compressor. This study develops two machine learning models to predict turbine and compressor decay, i.e. based on classification and regression approaches. The former classifies whether the components are decayed or not, thus highlighting a state of criticality, the latter predicts a specific value of each decay coefficient. For each approach, different algorithms are compared, e.g. boosted trees, linear regression or support vector machines. A case study considering sixteen inputs has been used to test the effectiveness of the proposed solution, starting from a dataset of about twelve thousand instances referred to a naval vessel. A sensitivity analysis of relevant parameters has been developed to verify the robustness of the approach.
2020
25th Summer School Francesco Turco
decay coefficients; naval propulsion system; condition-based maintenance; turbine; compressor; maritime maintenance
04 Pubblicazione in atti di convegno::04c Atto di convegno in rivista
Machine learning models to predict components decay in a naval propulsion system / Quatrini, E.; Colabianchi, S.; Costantino, F.; Tronci, M.. - In: ...SUMMER SCHOOL FRANCESCO TURCO. PROCEEDINGS. - ISSN 2283-8996. - (2020). (Intervento presentato al convegno 25th Summer School Francesco Turco tenutosi a Bergamo, Italy).
File allegati a questo prodotto
File Dimensione Formato  
Quatrini_Machine learning models_2020.pdf

solo gestori archivio

Tipologia: Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 497.75 kB
Formato Adobe PDF
497.75 kB Adobe PDF   Contatta l'autore

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1492089
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact