Högskolan Dalarnas logga och länk till högskolans webbplats

du.sePublikationer
Ändra sökning
RefereraExporteraLänk till posten
Permanent länk

Direktlänk
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • chicago-author-date
  • chicago-note-bibliography
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf
FAULT DETECTION IN AIR HANDLING UNIT (AHU) USING MACHINE LEARNING
Högskolan Dalarna, Institutionen för information och teknik, Mikrodataanalys.
Högskolan Dalarna, Institutionen för information och teknik, Mikrodataanalys.
2022 (Engelska)Självständigt arbete på avancerad nivå (masterexamen), 20 poäng / 30 hpStudentuppsats (Examensarbete)
Abstract [en]

Fault in Air Handling Unit (AHU) of the Heating, Ventilation, and Air Conditioning (HVAC) systems in buildings is a challenge that building managements face. These faults cause buildings to waste 15 – 30% of the energy consumed by the AHU. This thesis aims to study the causes of faults in the AHU and proposes a machine learning model that could be used to detect these faults. These faults could either be failure of AHU equipment, failure of the actuator, or failure of sensor and feedback controller. To achieve this, the data driven method of fault detection was applied. Collected data was preprocessed by removing missing values, eliminating correlated features, nominal features were one-hot encoded, and class imbalance was managed by over-sampling and under-sampling techniques. Finally, Principal Component Analysis (PCA) technique was applied to the over-sampled dataset. Both over-sampled and under- sampled datasets were split by 70/30 ratio for train and test sets, respectively. Classification models were built using Random Forest, Decision Tree and Support Vector Machines for both binary and multiclass classifications. GridSearchCV cross validation technique was used to train the models and the optimal parameters for each model selected. Results from multiclass classifications, show that Random Forest performed best with over sampling and PCA having 100% on accuracy, 100% on precision, 100% on recall, and 100% on f1 score while with under-sampling without PCA, Support Vector Machines performed best with 91% on accuracy, 91% on precision, 91% on recall, and 90% on f1 score. This illustrates that machine learning could be used to detect faults in AHU with accuracy above 90%. Analyzing the results, the proposed machine learning models could detect the most important failure causes and the predictors of faults in AHU.

Ort, förlag, år, upplaga, sidor
2022.
Nyckelord [en]
Fault Detection, AHU, Machine Learning
Nationell ämneskategori
Elektroteknik och elektronik
Identifikatorer
URN: urn:nbn:se:du-43907OAI: oai:DiVA.org:du-43907DiVA, id: diva2:1716261
Ämne / kurs
Mikrodataanalys
Tillgänglig från: 2022-12-05 Skapad: 2022-12-05

Open Access i DiVA

fulltext(2068 kB)701 nedladdningar
Filinformation
Filnamn FULLTEXT01.pdfFilstorlek 2068 kBChecksumma SHA-512
ebb1f21da77f99c962a855ef52cb6df5ba26c885398dc9f1ca6db2865a969afacc23c81db1c8a71e765b7a5d90ef6bf56c56f331e481e0a9b74340a839c70c78
Typ fulltextMimetyp application/pdf

Av organisationen
Mikrodataanalys
Elektroteknik och elektronik

Sök vidare utanför DiVA

GoogleGoogle Scholar
Totalt: 702 nedladdningar
Antalet nedladdningar är summan av nedladdningar för alla fulltexter. Det kan inkludera t.ex tidigare versioner som nu inte längre är tillgängliga.

urn-nbn

Altmetricpoäng

urn-nbn
Totalt: 630 träffar
RefereraExporteraLänk till posten
Permanent länk

Direktlänk
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • chicago-author-date
  • chicago-note-bibliography
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf