Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Fault classification in three-phase squirrel cage induction motor using artificial intelligence techniques

Ibrahim, Alasmer 2023. Fault classification in three-phase squirrel cage induction motor using artificial intelligence techniques. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of 2023Alasmer-PhD_Thisis.pdf]
Preview
PDF - Accepted Post-Print Version
Download (14MB) | Preview
[thumbnail of KEYWOR_1.PDF]
Preview
PDF - Supplemental Material
Download (94kB) | Preview
[thumbnail of Cardiff University Electronic Publication Form] PDF (Cardiff University Electronic Publication Form) - Supplemental Material
Restricted to Repository staff only

Download (337kB)

Abstract

Three-phase induction motor (IM) has various advantages in industries such as robustness, easy to control, high efficiency, and low cost. However, IMs are subjected to several faults during different operations. Different faults can cause losses in the production lines. Therefore, monitoring and diagnosing any faults in the motor during its operation is a very significant requirement to warn the operators so that any failure could be prevented. Condition monitoring and fault diagnosis approaches are needed to avoid any production loss as well as avoiding the high maintenance cost. Condition monitoring of IM is the process that checks and enables detection and prediction of different faults. The early fault diagnosis of the motor prior to a complete deterioration can provide an opportunity for maintenance to be performed and ensuring the safety without any problems in the industry. In this regard, the Artificial Intelligence (AI) techniques are considered as effective methods for the implementation of fault diagnosis of IMs. However, many challenges are still faced by the user of AI methods under the different real operative conditions which need to be solved before they can be implemented. The contribution of this research is the development of new hybrid fault detection model for three-phase induction motor using the AI approaches. The proposed model can achieve a higher classification accuracy with less training errors. Besides, this model is simple in design, fast to train, and reliable to test different size of input data. The suggested model consists of two components. The first component is suggested for feature extraction by applying Invasive Weed Optimization algorithm (IWO) and Convolutional Neural Network (CNN). While the second component is designed for classification task to process the different motor conditions using different algorithms such as k-Nearest Neighbour (KNN), Support Vector Machine (SVM), and Random Forest (RF). Different motor parameters including the stator current signal, motor vibration signal, and motor thermal images have been selected for investigation. Therefore, two applications have been proposed based on different feature extraction algorithms that can process the parameters of the motor manually and automatically. These approaches include firstly, the use of signal processing tools such as Matching Pursuit (MP), and Discrete Wavelet Transform (DWT) to process manually small size of input data that combined with the IWO algorithm based on feature selection technique, and secondly, another model was proposed based on the engagement of image processing techniques such Gramian Angular Fields algorithm (GAFs) with Convolutional Neural Network (CNN) to process big size of input data automatically. These approaches were successfully applied using MALAB R2021a and Python3.8 software, and the results proved their effectiveness. Measurements of some of the motor parameters were carried out by considering various single- and multi-electrical and/or mechanical faults. Three original sensor data were recorded simultaneously during laboratory experiments for investigation. Three common faults of induction motor were artificially generated in the lab for representing the motor conditions namely the rotor fault, the bearing fault, and the stator fault. The rotor fault includes three groups namely one Broken Rotor Bar (1BRB), five Broken Rotor Bars (5BRB)), and eight Broken Rotor Bars (8BRB). The bearing fault contains Ball Bearing (BB), Inner Bearing Race (IBR), and Outer Bearing Race (OBR)), lastly, stator fault is expressed as Open Circuit (OC) fault. The induction motor was assessed in the laboratory under the healthy and the faulty conditions. Then, the collected data was pre-processed using the analytical techniques to build an optimal feature matrix. In the first application, this data was processed manually in time-frequency domain to achieve the required optimum features. In other application, this data was converted to images by conducting a GAF transformation to obtain feature distributions. Then, the obtained data was fed to CNN to extract lower-level features. The assessment of the induction motor faults was achieved by calculating some evaluation metrics such as Specificity, Overall Accuracy, Precision, Sensitivity, and F1_Score. In addition, some efficient classification algorithms trained using k-fold cross-validation technique were selected to build the classification model. Different strategies have been adopted through this research to prove the efficiency of the suggested model. The achieved results showed that the proposed hybrid model was successful in its robustness for diagnosing the faults of the induction motor under different load conditions. This study has also shown that the combination of Artificial Intelligent techniques (AI) can be used as an optimum and reliable fault diagnosis model for other real-world fault detection applications.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Engineering
Uncontrolled Keywords: 1). Three-phase Induction Motor 2). Fault Classification 3). Artificial Intelligent Techniques 4). Invasive Weed Optimization Based Feature Selection 5). Transfer Learning 6). Deep Convolutional Neural Network
Date of First Compliant Deposit: 19 September 2023
Last Modified: 19 Sep 2024 01:30
URI: https://orca.cardiff.ac.uk/id/eprint/162472

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics