Show simple item record

dc.contributor.advisorLin, Xianke
dc.contributor.authorNwauche, Chukwuemeka Nelson
dc.date.accessioned2022-09-06T18:42:48Z
dc.date.available2022-09-06T18:42:48Z
dc.date.issued2022-08-01
dc.identifier.urihttps://hdl.handle.net/10155/1517
dc.description.abstractFault detection in lithium-ion batteries (LiB) is paramount to ensuring the long life and proper functioning of the batteries. To that end, this thesis proposes a combined fault diagnosis framework that leverages voltage charging curves and voltage charging curve fault residuals to accurately detect multiple faults within a LIB during partial and full charging regimes. This framework removes the need for parameter tuning and is also adaptable to varying battery chemistries and performs well with a small amount of available data. The framework leverages voltage residuals generated via a randomly initialized or pre-trained LSTM (Long Short Term Memory) model. Experimental results show its ability to accurately detect the different types of faults utilizing full voltage charging curve residuals with an accuracy of 95%. The framework can also detect faults utilizing partial voltage charging curve residuals & a pre-trained LSTM model with an accuracy of 94%.en
dc.description.sponsorshipUniversity of Ontario Institute of Technologyen
dc.language.isoenen
dc.subjectLithium-ion batteriesen
dc.subjectTransfer learningen
dc.subjectConvolutional neural networksen
dc.subjectElectric vehiclesen
dc.subjectBattery fault diagnosisen
dc.titleDeep transfer-learning based lithium-ion battery fault diagnosisen
dc.typeThesisen
dc.degree.levelMaster of Applied Science (MASc)en
dc.degree.disciplineAutomotive Engineeringen


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record