2024

Browse

Now showing 1 - 20 of 40

Time-Domain Modeling and Simulation of Performance Electric Vehicle Powertrain with Variable Battery Sizing Configurations
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-12) Islam, Md. Mahfuzul; Farazi, Shadman Saad; Hasan, Md. Syeed
Electric vehicle (EV) modeling has advanced to previously unheard-of levels of precision and sophistication in 2023 thanks to developments in simulation software, computing capacity, and the growing need for environmentally friendly transportation options. The state-of-the-art methods for EV modeling are examined in this work, with a particular emphasis on battery performance, powertrain efficiency, and dynamic vehicle behavior. This work presents a timedomain modeling and simulation framework for the Tesla Model Y's powertrain, aiming to enhance EV performance through precise simulation models. Key parameters such as mass, drag coefficient, rolling resistance, and wheel radius are integrated into MATLAB/Simulink. Proportional-Integral-Derivative (PID) controllers regulate motor current and vehicle speed, optimizing performance. The study explores variable battery sizing configurations, analyzing their impact on weight, internal resistance, acceleration, and efficiency. Results identify an optimal battery configuration, improving the balance between acceleration and power loss. This research contributes to advancing EV technology and sustainable transportation solutions.
Future Hybrid Energy System of Photovoltaic and Fuel Cell For Kanifing in The Gambia
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-13) Sanyang, Ousman; Ali, Akram; Ceesay, Lamin J; Bayo, Sarjo
The present research shows the outcomes of an optimal grid-connected with photovoltaic and fuel cell system design for Kanifing in the Gambia west Africa. The most efficient hybrid renewable power system is chosen by testing its performance and utilizing integrated modeling, simulation, optimization, and control methodologies. The key objective is to design a grid-connected with photovoltaic and fuel cell energy system with high utilization of clean energy, low greenhouse gas emissions, and a low cost of energy to meet the Kanifing's electric load. The performance and cost of the hybrid power system configurations using load executing and phase charging control techniques were assessed using hourly simulations, modeling, and optimization. Getting electricity is a significant difficulty in Africa. Although there is a lot of potential for using solar energy, there is little investment in renewable energy projects. Thus, a lot of people continue to rely on personal diesel generators, which emit large amounts of pollutants and harm both the environment and people. Situated in the Sunbelt, Gambia is one of the countries in Africa endowed with an extremely high solar irradiation potential. HOMER simulation software was used to determine the optimal configurations and sizes. A comparison is made between several hybrid combinations and a regular system. The studies showed that the suggested system had nearly lowered costs and CO2 emissions by 39% and 79%, respectively. The annual carbon footprint with avoided CO2 emissions is approximately 151,751 kg. The outcomes demonstrated that implementing a hybrid power system might be a reliable and profitable way to achieve social and environmental advantages in isolated rural and urban electrification
4G Data Analysis and Downlink Throughput Predictive Modelling Around The Geographic Region of Gazipur and Dhaka
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-22) Chowdhury, Shafi Muhtasim; Islam, Safwan; Islam, Rafrafin
This project presents a comprehensive study on optimizing 4G data analysis and downlink throughput modelling, integrating technical rigor with societal considerations. As 4G networks continue to serve as a backbone for modern telecommunications, enhancing their efficiency and performance is crucial for meeting growing data demands and supporting emerging applications. This project aims to develop advanced models and algorithms that improve network performance while addressing broader impacts on safety and societal aspects. The methodology involves the use of diverse resources, including datasets that were collected by hand from selected regions of interests in Gazipur and Dhaka, specialized software tools for model simulation, data analysis and network performance data collection, and case studies of real-world 4G network deployments. Advanced statistical methods, machine learning algorithms, and optimization techniques are employed to analyze data and model throughput. The project tries to emphasize the importance of data anonymization, security, and compliance with data protection regulations in an attempt to address ethical and privacy concerns. The results demonstrate that machine learning models are able to simulate and predict network downlink throughput with acceptable standards of accuracy, validated through rigorous analysis and evaluation. Detailed data analysis reveals patterns and trends that inform the optimization models, while comparative analysis with existing studies highlights the advancements achieved. In addition, this project underscores the role of engineering in society, addressing the ethical and societal implications of 4G technology. The findings contribute to the technical field of telecommunications while promoting sustainable and inclusive connectivity solutions.
Design of a Prototype of Smart Agriculture Powered by SPV
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-23) Siddo, Issaka Soumaila; Abubakar, Al-mustapha Babangida; Fari, Ahmad Mahmud
This project presents an IOT based smart irrigation system in order to optimize watering schedules for plants and lawns. The system uses an Arduino Uno microcontroller in conjunction with IoT(Internet of Things) sensors to measure soil moisture, temperature, and humidity, as well as a weather station to collect data on ambient conditions. The microcontroller processes this information and controls irrigation valves or motors to deliver water only when necessary. The system also includes a user interface for manual control and scheduling, and it can be connected to a network for remote monitoring and management. To further improve the sustainability and efficiency of the system, the project integrates a water fetching system that draws water from a well. This allows the system to supplement the water supply for irrigation and reduces reliance on external sources of water. In addition, the project incorporates a solar tracking system that maximizes energy generation from the solar panels to recharge the battery that powers the microcontroller. The results of the study demonstrate that the smart irrigation system significantly reduces water usage and enhances plant health when compared to traditional irrigation methods. This project presents a promising solution to the challenges of water scarcity and inefficient irrigation practices and has the potential to contribute to the development of more sustainable and environmen- tally friendly systems.
Time-based Scheme for Demand Side Management of ML-based Forecasted Load Data using an Optimization Algorithm
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-24) Anik, Mutasim Fuad; Jahan, A.S.M. Sarwar; Shamim, Md. Ashik Mia
This study focuses on developing a methodology to enhance the efficiency and sustainability of electrical grid operations through advanced load forecasting and demand-side management (DSM) strategies. This research aims to provide a robust solution to manage electricity consumption effectively, ensuring a balance between supply and demand. The study begins with an in-depth review of existing load forecasting methodologies, including traditional statistical approaches and modern machine learning techniques. Traditional methods, such as regression analysis and time series models (ARIMA), are discussed alongside more advanced techniques like neural networks and hybrid models, emphasizing their limitations and potential improvements. The research introduces a novel forecasting model integrating XGBoost with Particle Swarm Optimization (PSO) and an improved Long Short-Term Memory (LSTM) network. These models leverage comprehensive data from the PJM Hourly Energy Consumption dataset, providing a rich temporal coverage for accurate predictions. The dataset, spanning from 2002 to 2018, is meticulously pre-processed and split for training and testing to simulate real-world forecasting scenarios. The thesis also delves into various DSM strategies, including peak clipping, valley filling, load shifting, strategic conservation, and flexible load shaping. These techniques are essential for optimizing power consumption patterns, reducing operational costs, and enhancing environmental sustainability by minimizing carbon emissions. The implementation of these DSM strategies is analyzed in terms of their sociocultural, environmental, and ethical impacts. The study also highlights the benefits of incentivizing off-peak electricity usage, rate adjustments in achieving cost efficiency and reliable energy supply.
Cardiac Arrhythmia Detection by ECG Feature Extraction: A Machine Learning Based Approach
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Islam, Sadri; Ahsan, Auhona; Ahmad, Tanvir
ECG beats are vital for reducing fatalities from CVDs by enabling arrhythmia detection through intelligent systems, which provide crucial cardiac insights to specialists. However, challenges such as noise, heartbeat instability, and imbalance affect the accuracy and speed of these systems. Accurate diagnosis in quick time is essential for proper treatment and patient recovery. This study focuses on enhancing the precise diagnosis of various CVD types by analyzing arrhythmias in ECG signals of the heartbeats. We developed a deep learning based arrhythmia detection system that utilizes discrete wavelet transformation during pre-processing of the signals and the SMOTE oversampling algorithm to deal with the class imbalance problem. Our classifier integrates a Convolutional Neural Network (CNN) for spatial pattern detection with a Bidirectional Long Short-Term Memory (BLSTM) network for temporal dependency identification. We trained and evaluated our system using the MIT-BIH Arrhythmia Dataset.The evaluation results demonstrate that our method, after 50 training epochs, achieves high accuracy in different categories: 99.65% for class F, 99.37% for class V, 98.45% for class N, 99.10% for class S, and 99.82% for class Q. This proposed deep learning based system can be employed for the automatic diagnosis of arrhythmia and assist the CVD specialists in accurate diagnosis.
Conductive Ternary Nitride as an Alternative Material in the Design of Plasmonic Refractive Index Sensor
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Habiba, Anika Rahman; Rahman, Tasmiah; Prapti, Sumaiya Tasnim
Traditional plasmonic materials such as gold and silver have been widely used in Metal- Insulator-Metal (MIM) plasmonic sensors. However, these materials face significant challenges, including high optical losses, chemical reactivity and compatibility issues with standard fabrication processes. This thesis investigates the use of conductive ternary nitrides, specifically 𝑇𝑖(𝑥)𝑌(1−𝑥)𝑁, as a promising alternative for designing plasmonic refractive index sensors. Employing the Finite Element Method (FEM), the study examines the plasmonic properties of 𝑇𝑖(𝑥)𝑌(1−𝑥)𝑁, aiming to overcome the drawbacks of conventional materials. The initial parameters of the proposed sensor achieved a sensitivity of 513.3085 nm/RIU. After optimization, the sensor demonstrated enhanced sensitivity of 819 nm/RIU and an improved FOM of 32 RIU^(-1). The proposed plasmonic refractive index sensor, utilizing 𝑇𝑖(𝑥)𝑌(1−𝑥)𝑁, has demonstrated exceptional sensitivity in air pressure sensing applications, enabling precise detection of minute pressure changes. This capability is particularly beneficial for environmental monitoring and industrial applications where accurate air pressure measurements are crucial. The research provides a comprehensive comparison between 𝑇𝑖(𝑥)𝑌(1−𝑥)𝑁 and traditional plasmonic materials, underscoring the advantages of 𝑇𝑖(𝑥)𝑌(1−𝑥)𝑁, such as lower optical losses, higher chemical stability, and better compatibility with existing manufacturing technologies. The findings of this study pave the way for the development of efficient, high-performing plasmonic devices, advancing the fields of nanophotonics and plasmonic sensor technology.
Investigation on Mapping Highly Sensitive and Tunable Hollow Core Photonic Crystal Fiber Facilitated With Ultra-Short Pulse for Multipurpose Applications
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Nafiz, Abdullah Al Mahmud; Protiva, Afra Anika; Tamim, Mohammad
Over the years, researchers have proposed numerous designs for traditional PCFs, featuring various structures, sensitivities, and confinement losses. However, most of the designs proposed offer either high sensitivity with high confinement losses or low sensitivity with low losses and too much computations. After carefully reviewing several prior works and considering their problems, we have developed our sensors by analyzing changes in the shape of ultra-short pulses (USP) passed through Hollow Core Photonic Crystal Fiber (HC-PCF) using COMSOL Multiphysics 5.6. Nowadays, adulteration in fuel is a noteworthy concern due to its impact on engine performance, environmental pollution, and economic losses. Detecting adulteration in diesel fuel is a challenging task and it requires identifying adulterants without compromising safety or quality standards. We introduced a novel approach to sense diesel adulteration levels by analyzing changes in the shape of USPs passing through HC-PCF. This method advantages are fiber characteristics, including nonlinear parameters, to see how the shape of ultra-short pulses changes as they travel through diesel-filled HC-PCF. With the proposed sensor, we achieved remarkable compression sensitivity and power increases for diesel samples with varying adulteration levels under different input configurations. The method demonstrated a minimum sensitivity of 16%, indicating that the pulse is compressed by a factor of six, and the maximum power increase observed was 648.072 W.
Optimizing Stroke Risk Prediction Using Symptom-Based Feature Selection
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Bhuiyan, Refat Ahmed; Sarkar, Ahnaf Abid; Bisma, Tahya Ahammed
Stroke is a significant health concern, with early detection being challenging. Our research employs symptom-based feature selection using chi-square analysis and RFECV. By applying a logistic regression algorithm, we achieved 93% accuracy with just 9 features.
A Comprehensive Analysis on Network Slicing for Resource Allocation of 5G
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Akanda, A. G. M. Fuad Hasan; Khan, Rayan Hossain; Saqib, Shadman
Optimizing resource allocation in 5G networks involves reconciling the conflicting requirements of enhanced Mobile Broadband (eMBB), massive Machine Type Communications (mMTC), and Ultra-Reliable Low Latency Communications (URLLC). eMBB demands high data rates and substantial bandwidth to support applications such as high-definition video streaming and virtual reality. In contrast, mMTC requires the network to support a massive number of low-power, low-data-rate devices, essential for the Internet of Things (IoT). URLLC poses the additional challenge of requiring ultra-low latency and high reliability for critical applications such as autonomous driving and remote surgery. Advanced machine learning techniques, specifically Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, were utilized to develop prediction models. The CNN model achieved an impressive accuracy of 97% in predicting network slice allocations, while the LSTM model demonstrated a remarkable 97-98% accuracy in time series forecasting. Key achievements include enhanced model performance through meticulous hyperparameter tuning and data augmentation, which improved the model's robustness and generalization across diverse data scenarios. Processing time was significantly reduced by implementing early stopping and batch normalization techniques, accelerating model convergence and deployment. Additionally, optimized load scheduling ensured balanced workload distribution across the network, enhancing overall system performance and reducing latency. This comprehensive approach addresses the diverse and stringent demands of 5G services, demonstrating a robust, efficient, and scalable framework for 5G network resource allocation. This research ensures improved network performance and reliability, effectively meeting the varied requirements of eMBB, mMTC, and URLLC, thereby contributing to the advancement of next-generation wireless communications.
Fault Analysis in Electrical System using Machine Learning Algorithms
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Sakib, Md. Salman; Mahmud, Maleha; Shomyo, Md. Safin Mahmood
This research provides an overview of the prediction accuracy of different classification-based machine learning algorithms from a ‘Simulink based 3-phase Electrical System Model’. The model consists of a three-phase source and two relay bus bars connecting the ends of two subsystems. The data generation for faults was embedded by connecting a three-phase fault block between two subsystems. Later, the whole simulation was explained by interchanging the parameters among phase A, B, C and Ground. Each of the simulation placed between 0 to 1 second. The data table exhibits more than 16,000 samples across voltages ranging from -0.1023 V to 1.9776 V and a current ranging from -0.13813 A to 12.53622 for each of the faulty phase. More than 70% of the data was used for model training later and rest of the 30% raw data was planned for prediction. For classification, KNN, SVM, LR, DT, Gradient Boosting, Random Forest and MLP Classifiers algorithms were used for comparison. After several Preprocessing, ‘Random Forest’ algorithm comes with the highest accuracy of 0.9477 or 94.77%. Thus, the prediction can help the electrical engineers to automate the process of fault detection with the help of machine learning.
AI Trained IoT Based Automated Solar Panel Cleaning System
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Arika, All-Mumtahina; Fahim, Iffat Nowshin; Uddin, Jamal; Hossain, Salman
The collection of dust significantly decreases the efficiency of photovoltaic (PV) modules. In order to reduce the impact of dust on photovoltaic (PV) systems in a cost-efficient way, it is important to use optimal cleaning methods. The determination of the interval is required. In order to achieve this goal, machine learning (ML) models can be employed to identify the level of dust on photovoltaic (PV) systems that exceeds a predetermined threshold. This study aims to examine the effects of dust on photovoltaic (PV) systems in Bangladesh and suggests a new machine learning (ML) classification approach for detecting dust. Additionally, a cleaning system will be developed. Multiple machine learning classifiers were deployed and their performance was assessed. The Artificial Neural Network (ANN) emerged as the top-performing model, with an accuracy of 98.11%. When the machine learning model detects dust, the user can activate the water sprinkler cleaning system remotely. This technology successfully eliminates dust by spraying pressured water over the panel. The proposed cleaning mechanism successfully improved the efficiency of dusty PV modules to match that of clean modules (14.87%). A quantitative analysis was conducted to measure the reduction in productivity as a monetary loss in order to evaluate the feasibility of the cleaning system. The findings indicate that the suggested cleaning technique is financially feasible for photovoltaic systems with capacities above 2.89 kWp.
Enhancing Power Grid Reliability with Machine Learning Algorithms for Fault Detection and Classification
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Mobassir, Md. Sadi; Arpi, Safoat Saima; Fahad, Farhan-Uz-Zaman
In the complex and expansive networks of modern electric power systems, the occurrence of faults is an inevitable challenge that can significantly affect grid reliability and stability. This thesis presents a comprehensive study on the enhancement of power grid reliability through the application of machine learning algorithms for fault detection and classification. The primary focus is on the development and implementation of a hybrid model combining Long Short-Term Memory (LSTM) networks and Support Vector Machines (SVM) to accurately identify and classify various types of power system faults. The research begins with a detailed analysis of power system faults, including their causes, characteristics, and impacts on the stability of the grid. A significant portion of the study is dedicated to the classification of symmetrical and asymmetrical faults, with an emphasis on the most common and disruptive types such as line-to-ground and three-phase faults. The hybrid LSTM-SVM model is then introduced, highlighting its design, training, and validation processes. Empirical results demonstrate that the proposed model achieves high precision and recall rates across all fault types, with an overall accuracy of 96.7%. This high level of performance indicates the model's robustness and effectiveness in real-time fault detection and classification, making it a viable solution for practical deployment in power systems. Furthermore, the thesis integrates principles of Outcome-Based Education (OBE) to align the research with specific educational and professional outcomes. This approach ensures that the project not only addresses technical challenges but also enhances the competencies and skills of engineering students, preparing them for real-world applications and professional practices. The findings of this research contribute significantly to the field of electrical engineering by providing a robust methodology for improving power grid reliability. The successful application of machine learning techniques in fault detection and classification paves the way for further advancements in smart grid technologies and proactive fault management strategies.
Performance Analysis of Patch Antenna Sensors for Non-Invasive Body Electrolyte Monitoring
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Mahjabeen, Alisha; Ahmed, Raiyan Mustavi; Salsabil, Noshin
Non-invasive monitoring of electrolyte levels offers significant advantages over traditional blood-based tests, providing less discomfort and enabling continuous monitoring outside clinical settings.This thesis explores the use of microstrip patch antennas (MPAs) for non-invasive electrolyte sensing. It examines three MPA designs: a simple microstrip patch, a spiral engraved sensor patch, and a patch antenna with T-shaped slots. These designs were modeled and analyzed using COMSOL Multiphysics simulations to assess their performance in detecting varying concentrations of sodium chloride (NaCl) in sweat. Key performance metrics such as sensitivity, accuracy, and precision were evaluated based on the reflection coefficient (S11 parameter). The results show that each antenna design has unique advantages and limitations regarding sensitivity to electrolyte changes and practical integration into wearable devices. Innovations in antenna design, such as the RFID-inspired spiral engraved patch and T-slotted patch antennas, show promise in enhancing sensitivity and user comfort for continuous health monitoring. Despite advancements, challenges like environmental interference and the need for greater sensitivity to small biological changes remain. The study shows that the T-slotted patch antenna with a barium titanate slab achieved top sensitivity, precision, and accuracy for detecting NaCl levels in sweat, despite cost and safety concerns. It had the highest accuracy (95.62%), while the spiral engraved sensor patch model excelled in precision with a 0.0026 standard deviation and 7. 00 × 10 variance. The simple −6 microstrip patch antenna offers a cost-effective alternative with 95.51% accuracy. The study highlights the ongoing need for innovative antenna designs to overcome these challenges, ensuring that non-invasive electrolyte sensors can be effectively utilized in healthcare monitoring. This research identifies optimal design parameters for MPAs to enhance non-invasive electrolyte sensing, aiming to advance technology and improve integration into next-generation medical devices.
Design of a Highly Sensitive Photonic Crystal Fibre Sensor for Detecting Biochemical Analytes
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Dola, Fariba Tabassum; Awsaf, S M Azmain; Apon, Jubair Mahamud
We proffer in this research a distinctive, facile to fabricate, and highly sensitive photonic crystal fiber (PCF) biosensor based on the phenomenon of surface plasmon resonance (SPR). Our prototype has a strategic pattern of circular air holes inside the fiber, which leads to a superior sensing performance. The evaluation of all the sensor characteristics has been discharged by employing the finite element method (FEM) of COMSOL Multiphysics. The gold (Au) layer just around the fiber acts as the plasmonic material. After the optimization of all the fiber parameters, we derived a maximum amplitude sensitivity (AS) and wavelength sensitivity (WS) of 2202.64 RIU− 1 and 140,500 nm/RIU, respectively, with a maximum sensor resolution 7.11 × 10− 7 for wavelength and 4.54 × 10− 4 for amplitude. Moreover, the maximum figure of merit (FOM) procured was 2285. The overall analyte sensing range is from refractive indices 1.31 to 1.40, and the sensor has a fabrication tolerance limit of ±5% for the gold layer variation and ±2.5% for both of the air holes. With its enhanced performance in terms of sensitivity, we believe that this SPR-based PCF biosensor can potentially contribute to the detection of unknown analytes and in applications of medical diagnostics.
Novel CMOS-Compatible Plasmonic Pressure Sensor with Silicon-Insulator-Silicon Waveguide Configuration
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Kabir, Mohammad Abrar; Keats, Aseer Imad; Taharat, Abdullah
This thesis introduces a novel CMOS-compatible plasmonic optical pressure sensor featuring a Silicon-Insulator-Silicon waveguide configuration. The sensor design incorporates a Railtrack resonator coupled to a straight waveguide with gratings, further enhanced by embedding silicon nanorods into the resonator cavity. This sensor demonstrates a notable redshift in the transmission spectrum related to the deformation of the resonator structure under applied pressure. The proposed sensor exhibits an unprecedented pressure sensitivity of 51.075 nm/MPa, arguably the highest value reported to date for Metal-Insulator-Metal based pressure sensors. Moreover, this work represents a novel instance of employing CMOScompatible silicon for designing an optical pressure sensor, thereby bridging the gap between plasmonic optomechanical sensors and nanoelectronics, while circumventing the compatibility issues typically associated with metals in standard CMOS fabrication processes. While traditional metals suffer from limited tunability due to their inherent carrier concentration constraints, silicon offers a promising solution as its optical properties can be finely tuned by modifying the doping levels, addressing the challenge of optical tunability. The sensor’s versatility and impact across diverse domains are highlighted by its potential applications, including gas leakage detection, flow rate measurement, electronic skin sensing, and pressure sensors as refractive index sensors for early diagnosis of organ rejection post-transplantation.
Optimized Plasmonic On-Chip Refractive Index Sensor for Biosensing Applications
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-25) Rahman, Asma; Ishrat, Samiha; Abdullah
This thesis focuses on designing plasmonic nanosensors for high-sensitivity detection of molecular interactions at the nanoscale, using SPPs and light-matter interaction. Two sensor designs, utilizing silver and ZrN materials, demonstrate significant sensitivity improvements via FEM analysis and Drude-Lorentz Model. Applications include medical diagnostics (e.g., cancer cell classification, detection of anemia and diabetes). The use of ZrN enhances sensor performance due to its exceptional optical and electrical properties, making it compatible with CMOS technology. OBE requirements are discussed regarding health, socio-cultural factors, and environmental sustainability.
Printed Circuit Board Defect Detection Using Convolutional Neural Networks
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-26) Rashid, Ahmed Jawad; Isfara, Adiba; Ullah, Mohammad Aman
The increasing complexity and miniaturization of modern electronic devices necessitate highly reliable and defect-free Printed Circuit Boards (PCBs). Effective defect detection in PCBs is crucial to maintaining the quality and reliability of these devices. However, current PCB defect detection datasets exhibit significant limitations that hinder the development of robust and accurate models. Existing datasets are limited in scope, do not accurately mimic real-world defects, and fail to represent the diversity and complexity of industrial PCBs. For instance, PKU dataset is restricted to only a small amount of PCB boards and include defects introduced post-manufacturing using Photo Editing Applications, which do not closely mirror real-world manufacturing imperfections. Additionally, these datasets label each PCB board with only one fault class, despite real-life scenarios where multiple faults can occur simultaneously. These compounded limitations make the datasets less suitable for generalizing to the diverse and complex situations encountered in real-world PCB inspections. To address these challenges, this research aims to develop a comprehensive and realistic dataset created through chemical etching procedures, reflecting the true nature of manufacturing imperfections. Additionally, we trained advanced Convolutional Neural Network (CNN) models, including YOLOv8, HRNet, Cascade R-CNN, ATSS, RetinaNet, and Faster R-CNN, to detect six common PCB defects: missing pad, open circuit, short circuit, spur, spurious copper, and mouse bite. Our findings indicate that YOLOv8 demonstrated superior accuracy and speed, achieving a mean Average Precision (mAP) of 0.888 at 50% intersection over union (mAP50) and a detection speed of 16.3 frames per second (FPS). HRNet, while achieving the highest mAP50 of 0.905, was less suitable for real-time applications due to its lower frame rate of 9.2 FPS. ATSS and Cascade R-CNN offered balanced performance with mAP50s of 0.88 and detection speeds of 14.51 FPS. In contrast, RetinaNet and Faster R-CNN were less effective due to lower accuracy and slower processing times. This study underscores the inadequacies of existing PCB datasets and the necessity for more accurate and efficient defect detection models.
Hysteresis Characteristics Analysis for Ferroelectric TFTS and It’s Application to Dram
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-26) Mahmud, Shah Md. Ashik; Islam, Shadid; Wornob, Shafaiet Newaz; Basher, Md.Khairul
This thesis presents a brief analysis of the hysteresis characteristics of various ferroelectric materials. Ferroelectric materials shows spontaneous remanent polarization that can be modified by an external field. This property is crucial for applications in non-volatile memories, actuators, and sensors. By examining hysteresis loops, the study aims to understand the distinct behavior of ferroelectric materials under different electric fields and their impact on practical applications. Experimental and theoretical approaches are used to elucidate the mechanisms driving hysteresis in ferroelectrics, providing insights into optimizing material performance for specific applications.
Low Frequency Noise Modeling of Ferroelectric Thin Film Transistor
(Department of Electrical and Elecrtonics Engineering(EEE), Islamic University of Technology(IUT), Board Bazar, Gazipur-1704, Bangladesh, 2024-06-26) Shaf, Mohd Muztahid Abrar Siddique; Shams, Liman; Ahmed, Safwan
Low-frequency noise (LFN) modeling and characterization in ferroelectric thin-film transistors (FeTFTs) is crucial for understanding and mitigating the effects of noise on device performance. This paper presents a comprehensive approach to LFN modeling that incorporates the unique characteristics of ferroelectric materials. Traditional TFT models are extended to include polarization-induced hysteresis, providing a more accurate representation of FeTFT behavior. Enhanced noise integration techniques are employed to account for various sources of low-frequency noise, including 1/f noise and thermal fluctuations. Parameter refinement for off current modeling is conducted to improve precision in predicting leakage currents. Additionally, advanced dynamic response models are developed to capture the timing and responsiveness of drain current under varying operational conditions. The convergence of memory window characteristics at peak gate voltages is improved to ensure reliable device operation. This work also refines the transfer characteristics equations to better reflect the low-frequency noise inherent to FeTFTs. Experimental results validate the proposed models, demonstrating significant improvements in noise characterization and device performance prediction. The findings contribute to the development of more reliable and efficient FeTFTs, suitable for applications in memory storage and low-power electronics. This thesis focuses on characterizing FeTFT, modeling the noise, as well as developing a Graphical User Interface (GUI) to simulate the memory window of the device.