Pusat Pengajian Kejuruteraaan Elektrik dan Elektronik - Tesis
Browse
Recent Submissions
Now showing 1 - 5 of 528
- PublicationFuzzy-based vector control system for permanent magnet synchronous motor(2024-12-01)Zhou, HongqiangPermanent Magnet Synchronous Motor (PMSM) is highly regarded for its efficiency, power density, and dynamic performance, making it integral in industrial automation and electric vehicles. However, traditional PI controllers often struggle with parameter variations, nonlinear characteristics, and load disturbances, affecting steady-state and dynamic performance. This thesis addresses these challenges by proposing a Fuzzy PI control strategy that combines Field-Oriented Control (FOC) with Fuzzy logic, enabling adaptive adjustments of PI parameters for enhanced stability, robustness, and responsiveness. MATLAB/Simulink simulations show that the Fuzzy PI controller achieves faster steady-state response, reduces steady-state error, and minimizes overshoot across various speeds, including 500 and 1200 r/min. During a load disturbance test with a drop from 50% (5 Nm) to zero at 2.5 seconds, both controllers reached a similar overshoot peak (about 600 rpm); however, the Fuzzy PI controller stabilized back to 500 rpm within 0.3 seconds, significantly faster than the 0.8 seconds of the conventional PI controller. Additionally, the Fuzzy PI controller reduced current oscillations (±4 A versus ±6 A), further affirming its superior dynamic response and precision. These results validate the Fuzzy PI control strategy's effectiveness in optimizing PMSM performance, offering a robust framework for advanced control applications.
- PublicationMotor fault classification using thermal imaging and modified inceptionv3 model(2024-12-01)Xu, LifuIndustrial motors are considered one of the essential equipment widely used in various sectors. However, due to factors such as prolonged operation, environmental conditions, and inadequate maintenance, industrial motors are prone to various failures. This study proposes a thermography-based motor fault detection method utilizing the InceptionV3 model, addressing its limitations in handling noise, low-contrast images, and small datasets through several enhancements. To overcome the problem of low contrast in thermal images, Contrast Limited Adaptive Histogram Equalization (CLAHE) is applied to the input images. Furthermore, the Squeeze-and-Excitation (SE) channel attention mechanism is integrated into the InceptionV3 model to improve its performance. The proposed model was tested using a publicly available dataset containing 369 thermal images of an electric motor with 11 types of faults. Experiment results show that an accuracy of 98.13% is achieved by the model. The trained InceptionV3-SE model was used for feature extraction to train a Support Vector Machine (SVM) classifier, which attained a maximum accuracy of 100%. Achieving perfect accuracy in this context highlights the proposed method's ability to overcome challenges such as class imbalance and blurred inter-class boundaries, setting a new benchmark for motor fault detection systems. This research contributes to the field of industrial motor fault classification. The effectiveness of the proposed method in accurately identifying various motor faults is demonstrated, which holds significant value for real-world industrial applications.
- PublicationPhysical modeling of non-alloyed ohmic contact towards gallium nitride-based high electron mobility transistor applications(2024-10-01)Tung, Kok SiongAdvantages of GaN's High Electron Mobility Transistor (GaN HEMT), such as concentrated channel electron, superior electron mobility characteristic and high breakdown voltage bring the opportunity to replace the Silicon-based devices in the near future of modern power conversion systems. Low resistance Ohmic contacts of AlGaN/GaN-based devices are essential to achieve forecasted device performance. There hasn’t been much work done on the device’s software based system (TCAD) to study the contact resistance for GaN and establish a strategy to minimize the contact resistivity. This paper objective was to study the Ohmic characteristic of metal contact on HEMT semiconductor with perform simulation with develop a physical mode reflecting it current transport mechanism and extract the TLM plot from I-V curve go gain the contact resistivity. In this research work, using Silvaco TCAD Atlas, the study started with modeling the contact resistance with vertical structure. The study extended to the lateral structure, which is more feasible for physical manufacturing, whereby different n++ with various doping under metal were studied to obtain the best optimization for the ohmic contact. Increasing the doping of a semiconductor, resulted in higher possibility of tunneling as the width of barrier become narrower, current flow in the form of field emission cause the reduction of resistance. Base on vertical structure simulations, sheet resistance value obtain from TLM plot compare to theoretical calculations using formula of s = 1/qμNd are matching proven that the validity of the model. Base on the lateral structure simulation TLM plot, it can be seen that when doping rises, the slope representing the sheet resistance decreases even the mobility of electron or hole increase infer that increasing of doping overcome the effect of scattering or collision. In n++ under metal structure, the heavily doped layer enable more electron tunneling at the junction with higher doping and higher heavily doped layer thickness which enable higher possibility of tunneling. Hence, higher current density able the cross the metal-semiconductor junction with lower transfer length and thus lower the contact resistivity. Based the the data extracted from the IV curved and TLM plot, reduction in contact resistivity saturated after 18 nm thickness and contact resistivity achieve < ~1E-6 Ω∙cm2. TLM parameters are in good agreement with the theoretical sheet resistance which demonstrate that the validity of the model. It also revealed that with n++ layer under metal Ohmic contact was observed, in contrary Schottky contact was observed without n++ layer under metal. However, simulations indicate a significant of higher contact resistivity than the experimental values. This is because the simulations are performed under perfect conditions, with no surface defects or recombination centres. Also, the flaws, traps and surface contaminants was not consider in the simulations are to be blamed that the differences between simulations and actual devices. It is concluded from this study that a heavily doped metal layer exists in the metal-semiconductor interface which enable the metal contact to form Ohmic contact.
- PublicationOptimal dispatching and flexible topology methods for urban power systems(2024-01)Su, YiThis thesis aims to investigate the optimal operation of three forms of Urban Power Systems (UPSs) which will be co-existence in a long time. First of all, aiming at AC / DC hybrid distribution system, the thesis proposes a comprehensive optimal dispatching model with centralized two-stage dispatching framework, achieving a 25.9% reduction in power loss compared to the initial model and increasing the integration of renewable distributed generations from 71% to 200%. Secondly, aiming at AC / DC hybrid active distribution system, the thesis incorporates social behaviours of electric vehicles and power demand response within a three-stage energy management framework, leading to a 13.07% enhancement in social welfare and a 19.90% increase in calculation speed. At last, aiming at transmission and distribution coupling system, the thesis proposes a two-layer optimal dispatching framework that takes into account the switching sequence along with dynamic thermal rating, leading to the effective resolution of congestion issues across various scenarios while maintaining a safety margin. In general, the models and optimization dispatching framework proposed in this thesis can help achieve the optimal power flow for UPSs, and improve its economic, security, and stable operation.
- PublicationChinese traffic sign detection and recognition based on lightweight you only look once (YOLO) models(2024-08)Song, Wei ZhenDetecting and recognizing traffic signs is crucial for intelligent driving systems, providing essential real-time guidance to drivers. Challenges such as bad weather, lighting, and occlusions hinder traffic sign detection. Conventional algorithms struggle to balance accuracy and real-time performance, leading to the favouring of lightweight deep learning detection algorithms for their automatic feature extraction and low computational cost. This study is based on the classic YOLOv4-tiny and YOLOv5s object detection algorithms, proposing several improvement strategies aimed at developing a more robust model for detecting traffic signs. This research selects the Tsinghua-Tencent 100K (TT100K) dataset and the CSUST Chinese Traffic Sign Detection Benchmark (CCTSDB and CCTSDB2021) datasets for training and evaluating traffic sign detection algorithms. Enhancements include an improved lightweight Better Efficient Channel Attention (BECA) mechanism, an upgraded Dense Spatial Pyramid Pooling (Dense SPP) network, an extra detection head, and optimized anchor boxes. The improved TSR-YOLO model showed significant improvements in precision (96.62%), recall (79.73%), F-1 Score (87.37%), and mAP (92.72%) with a stable FPS of around 81. However, due to its complexity, it is unsuitable for embedded devices. Thus, the study developed Sign-YOLO, which has been improved by a Coordinate Attention (CA) module, a High Bidirectional Feature Pyramid Network (High-BiFPN), and the Better Ghost Module to reduce model size. Sign-YOLO was evaluated on the CCTSDB2021 and TT100K datasets, reducing parameters by 0.13M compared to YOLOv5s, achieving a good balance between accuracy and speed for traffic sign detection.