hig.sePublikasjoner
Endre søk
RefereraExporteraLink to record
Permanent link

Direct link
Referera
Referensformat
  • apa
  • harvard-cite-them-right
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annet format
Fler format
Språk
  • sv-SE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • de-DE
  • Annet språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf
Prediction of industrial machine failure by analysing anomalies
Högskolan i Gävle, Akademin för teknik och miljö, Avdelningen för elektroteknik, matematik och naturvetenskap.
2022 (engelsk)Independent thesis Advanced level (degree of Master (Two Years)), 20 poäng / 30 hpOppgave
Abstract [en]

The sudden downtime and unplanned maintenance not only drastically increase the maintenance cost but also decreases the production capacity for the manufacturer industries. This is because the machines on these industries fail suddenly and totally stop the production as the machine should be fixed by maintenance before it can run again. To deal with it, several maintenance techniques have been adopted. But as soon as an automated maintenance technique comes in named predictive maintenance, the future machine failure can be predicted. To perform this prediction, a synthetic dataset is used that is taken from 100 industrial machines. From this dataset, the simulated sensor data, error, and failure history have used to calculate the probability of error and failure during the time period of an anomaly. This probability is calculated by the basic probability equation. In addition, the sum of the calculated probability of error and failure, give the intuition about the most relevant sensor data for a machine. This relevant sensor data is then used as response for the prediction with gaussian process regression algorithm. This prediction of response is shown for machine number 85 which is the most important from all 100 machines as this machine is very sensitive to any of the 4 sensor anomalies. Then, the sum of probability can be coherent with the anomaly on the predicted response which is the most relevant sensor data. This defines that the machine is in high risk of experiencing machine failure and thus the machine should be fixed by adopting maintenance. In contrast, the opposite is also true for low probability of error and failure for an anomaly on the predicted response. To evaluate the performance of the algorithm, four statistical metrics are used among which three matric is to estimate the errors and the other one is the correlation coefficient between the actual and predicted data. 

sted, utgiver, år, opplag, sider
2022. , s. 36
HSV kategori
Identifikatorer
URN: urn:nbn:se:hig:diva-38260OAI: oai:DiVA.org:hig-38260DiVA, id: diva2:1647091
Fag / kurs
Electronics
Utdanningsprogram
Electronics/Automation – master's programme (two years) (sv or eng)
Veileder
Examiner
Tilgjengelig fra: 2022-03-25 Laget: 2022-03-24 Sist oppdatert: 2025-10-02bibliografisk kontrollert

Open Access i DiVA

fulltext(2139 kB)474 nedlastinger
Filinformasjon
Fil FULLTEXT01.pdfFilstørrelse 2139 kBChecksum SHA-512
58751c7354dae000d7dd06d73a6c08445cfd1284fb7bf44841eabfb6b558d7dbf0140f90f4b1d6048bbb4e786176f4729820b2d0e33bf1b56f74401a598e66aa
Type fulltextMimetype application/pdf

Av organisasjonen

Søk utenfor DiVA

GoogleGoogle Scholar
Totalt: 475 nedlastinger
Antall nedlastinger er summen av alle nedlastinger av alle fulltekster. Det kan for eksempel være tidligere versjoner som er ikke lenger tilgjengelige

urn-nbn

Altmetric

urn-nbn
Totalt: 954 treff
RefereraExporteraLink to record
Permanent link

Direct link
Referera
Referensformat
  • apa
  • harvard-cite-them-right
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annet format
Fler format
Språk
  • sv-SE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • de-DE
  • Annet språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf