aalto1 untyped-item.component.html
Sensor fusion based condition monitoring of induction motor using artificial intelligence techniques
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
Department
Major/Subject
Mcode
ELEC3024
Language
en
Pages
49+11
Series
Abstract
Industries are proliferating, and the need for induction motors (IMs) plays an essential role in various technical and economic fields. Due to the excessive usage, several faults occur in the machine. It is imperative to detect faults at an early stage. Condition monitoring techniques can detect unexpected failures in the machine. Otherwise, they may cause massive losses for industries. Several methods have been investigated using a single sensor and signal-processing methods. However, using a single sensor does not facilitate in detecting different types of faults. Therefore, several current and vibrations sensors are used to identify various forms of defects. However, data derived from this approach is too vast and complicated. A number of artificial intelligence technologies are being used to solve this. The support vector machine (SVM) has been used for this thesis due to its excellent classification efficiency compared to other techniques.
The main focus of the thesis is on the development of a sensor fusion-based approach for condition monitoring of IMs using SVM. This was achieved by adequately preprocessing the dataset, extracting the maximum number of features, choosing the suitable features, identifying the hyperparameters and designing the model. The thesis investigates three main processes: binary classification, multiclass classification, and sensor fusion. The algorithm's working condition has been checked in binary classification, and feature selection methods have been evaluated. The accuracy of different classes is monitored using a single sensor in a multiclass classification method. The sensor fusion approach has been used to test accuracy changes with the fusion process at the feature level. The fused multi-dimensional information is used to train a multiclass SVM with eight classes. Finally, the proposed approach is tested by current and vibration data. The results prove that the feature level fusion method has great potential in increasing the accuracy of fault diagnosis.