Pusat Pengajian Kejuruteraaan Elektrik dan Elektronik - Monograf

Browse

Recent Submissions

Now showing 1 - 5 of 968
  • No Thumbnail Available
    Publication
    Fault detection of electrical motor based on thermal imaging and machine learning
    (2023-08)
    Yap, Jun Hong
    Faults could occur on electrical motor due to various reasons, and an early motor fault detection system helps prevent interruption in service and financial losses. However, the current practice of manual fault inspection and preventive maintenance is time consuming, and it may not be effective. Thus, motor fault diagnosis using thermal imaging technique has been on the rise in recent years. To further improve the effectiveness and to automate fault detection using thermal imaging, artificial intelligence (AI) can be employed. Hence, in this project, an electrical motor fault detection system based on thermal imaging and machine learning (ML) technique was developed. Transfer learning (TL) approach using pre-trained convolutional neural networks (CNNs) was used. The CNN was trained to learn the features extracted from the thermal images of a faulty and a healthy motor and use them to diagnose the condition of the motor. Various hyperparameters were configured for network training to obtain the best results. Furthermore, performance analysis was conducted and discussed to evaluate the credibility and reliability of the trained network. A Graphical User Interface (GUI) was then created to ease the user in using the proposed fault detection system by just supplying the thermal images of a test motor to the GUI application for fault diagnosis. The evaluation results showed that GoogLeNet gives the best detection performance with both the mini-batch and the validation accuracy achieving a 100%, and both the losses were low as well, at 0.0015 and 0.0001 respectively. Thus, the final trained network based on GoogLeNet was used in the GUI for the implementation of the proposed motor fault detection system. In conclusion, the aim for implementing a fault detection system and GUI, through the use of thermal images and machine learning was achieved.
  • No Thumbnail Available
    Publication
    Explainable artificial intelligence for signature verification system
    (2023-10)
    Lee, Sze Yuan
    In recent years, the use of personal identity, such as signatures, as a means of authentication has gained significant attention. There are some concerns arise due to the potential for signature forgery and leading to the development of signature verification systems to determine the authenticity of signatures. The lack of understanding behind the AI and DL can erode trust in the tools as incorrect or biased decisions made. The application of Explainable Artificial Intelligence (XAI) methods in signature verification systems can address these concerns by providing insights into the decision-making process and enhancing the trustworthiness and reliability of the system. This research aims to explore and evaluate various explanation models to improve the interpretability and performance of signature verification systems. Furthermore, this research seeks to identify the specific aspects that users and developers focus on when considering explanations generated by these models. Moreover, this research aims to develop a new explanation model by combining the strengths of two widely used methods, LIME and Grad-CAM. The experiment is conducted through MATLAB using package known as Deep Learning Toolbox. The explanation evaluates through the respond of 18 respondents in four aspects, understandability, interpretability, accuracy and usefulness. The survey is also used to identify the evaluation aspect that are focused by users and developers. In addition, a new explanation model is developed through the combination of “scoremap” of LIME and Grad-CAM. Preliminary findings indicate that the Grad-CAM method demonstrates better performance from the user's perspective, while developers tend to prefer the LIME method. By leveraging the strengths of both approaches, the new explanation model achieves an impressive increase in understandability, interpretability, accuracy, and usefulness.
  • No Thumbnail Available
    Publication
    Evaluation of isolated type multilevel inverter with different DC source selection scheme
    (2023-07)
    Ooi, Wei Tao
    Multilevel inverters (MLI) play a major role in various power applications in converting direct current to alternating current for a power system. The ideal output voltage waveform should be a perfect sine wave. In practical, a real inverter will produce output voltage waveform which contains harmonics that are not acceptable for high power applications. Harmonics cause current and voltage waveforms to be distorted, resulting in power system degradation. To date, reducing the total harmonics distortion (THD) is always the main objective to conduct the research regarding the multilevel inverters, especially when the inverters are applied in renewable energy application. Therefore, the simulation testing of this project will focus on the total harmonic distortion value of the system. The proposed multilevel inverter in this research is asymmetrical multilevel inverter with trinary sequence. The targeted number of output levels are 5-L, 7-L and 9-L. This research is aimed to analyze the proposed topologies from the aspects of inverter power ratio, total harmonic distortion level and power contribution of DC voltage source in different DC source scheme. The switching signals are generated using a low frequency modulation technique where the switching angles are pre-calculated using derived mathematical equations. To evaluate the performance of the topology, linear load tests are conducted to ensure proper operation of the proposed topology in reference to theoretical analysis with MATLAB Simulink software. Based on simulation result, the THD values of the 5-L waveforms are the highest, followed by the 7-L waveforms and the lowest THD are measured on the 9-L waveforms. Under RL loads, the current waveforms show much lower THD readings since they are filtered by the inductive loads. The RL filtering effect causes the current waveforms to lost the stepped pattern which explains their smoother waveforms. The results show the high power ratio is averagely achieving 95% and above, indicating the reliability of the system. From DC source power contribution aspect, it is concluded that high power source is required to produce higher number of levels in MLI inverter. In short, the research outcome shows the topology with different DC source configuration is capable of generating high number of output levels with low number of total components. The individual current harmonics are also in accordance to the IEC/EN 61000-3-2 standard. Hence, this topology has great potential to be developed for real applications since the production cost is expected to be lower than the other topologies.
  • No Thumbnail Available
    Publication
    Evaluation of conventional and modified hill climbing mppt algorithms
    (2023-08)
    Siti Khadijah binti Shamsulkamal
    The conventional perturb and observe (P&O) algorithm, one of the methods of Hill Climbing maximum power point tracking (MPPT), is well-liked due to its simplicity of implementation. Nonetheless, conventional P&O has limitations because of power oscillations in the steady state and poor tracking of MPP with inadequate choices of perturbation step size and tracking time. Additionally, even after achieving a steady state, continual perturbations still result in wasteful power oscillations. On the other hand, a significant study has been conducted on the P&O MPPT methods for enhancing the overall effectiveness of solar photovoltaic (PV) systems. In this paper, an evaluation of the conventional and the recently published modified P&O hill climbing MPPT algorithm approaches is proposed to solve these drawbacks of the conventional P&O algorithm. By modelling the three algorithms in MATLAB Simulink and comparing their performance to that of the conventional P&O algorithm, the effectiveness of the three-enhanced method has been confirmed. It can be seen that the Modified 1 P&O method surpasses the Modified 2 and 3 P&O MPPT methods as well as conventional P&O. The Modified 1 P&O algorithm is the most effective way to lower steady-state oscillation since it has a low average input power ripple (0.19336%) and the lowest average output voltage ripple (0.05602%). In comparison to other approaches, it also offers the fastest simulation time, the shortest tracking time, and the highest MPPT efficiency (99.97%).
  • No Thumbnail Available
    Publication
    Enabling data visualization through an iot and cloud-based monitoring platform
    (2023-08)
    Muhammad Danial bin Mahamad Farizan
    IoT represents the integration of multiple technologies, such as cybersecurity, cloud computing, big data, and artificial intelligence. The IoT enables connection between each device and sensors. The implementation of IoT technologies can lead to increase productivity and efficiency. This project aims to explore the use of IoT technologies with the development of a cloud-based IoT platform for data visualization. The cloud-based platform is designed to improve the efficiency of the IoT technologies at an affordable cost. The platform will allow the user to monitor data and receive a notification based on the data received from the device