Electrical & Computer Engineering Masters Theses

Permanent URI for this collectionhdl:10365/32560

Browse

Now showing 1 - 20 of 84

Automating the design of deep learning models using neural architecture search for medical image classification
(North Dakota State University, 2024) Balamurugan, Ridhanya
Designing Deep Learning (DL) models for medical image classification tasks poses significant challenges, demanding substantial expertise owing to the intricate nature and critical importance of the undertaking. Creating a DL model tailored for such purposes entails iterative processes of designing, implementing, and fine-tuning algorithms to achieve optimal performance. To mitigate these difficulties, Neural Architecture Search (NAS) has risen as a key field to generate the most effective DL models automatically. However, much of the previous studies involving NAS focus on automating the design of DL models for well-established datasets such as CIFAR-10 and ImageNet. This technique should also be extended to medical image datasets where detecting crucial features accurately in medical images is essential to detect specific illnesses correctly. Therefore, in this study, we investigate NAS to autonomously design best performing DL models for skin lesion detection, thereby demonstrating its usefulness for additional medical image classification endeavours.
The integration of multiple cardiovascular and pulmonary models into a single unified cardiopulmonary model
(North Dakota State University, 2024) Thrash, Julian III
There has been an increasing effort to employ cardiovascular (CV) and pulmonary models to assist in the study of disease, answer research questions, and study the effects of a medical treatment or device. To properly study the intricate dynamics of the CV and pulmonary systems, a comprehensive model is required to achieve a holistic understanding of its mechanics and view the emergent properties of the complex and dynamic cardiopulmonary system. This study builds on the work of Mauro Ursino et al., by taking their multiple CV and pulmonary models and integrating them together within the scalable environment of MATLAB and Simulink. Multiple published mathematical models were implemented and integrated together within Simulink and MATLAB, its hemodynamics and lung mechanics were verified, and the model was additionally adapted for use simulating an invasive pressure-volume (PV) study of the effects of vascular aging of the aorta on cardiac function and efficiency.
Large spacecraft medium voltage DC (MVDC) power and propulsion system
(North Dakota State University, 2024) Talebzadeh, Sarah
This thesis introduces a medium voltage direct-current (MVDC) system for large spacecrafts with megawatt (MW)-scale power and propulsion systems for interplanetary transport. The proposed MVDC system comprises a nuclear electric propulsion (NEP) powered permanent magnet generator, a solar photovoltaic (PV), a backup battery energy storage system (BESS), an electric thruster, AC/DC loads, and the spacecraft low-voltage DC (LVDC) power system. The NEP is the main power source for the thruster, while the solar PV and BESS support payload and low-voltage systems. The energy sources are connected to the MVDC bus through power electronics converters. Unidirectional boost DC-DC converter and bidirectional boost DC-DC converters are adopted on the PV and BESS sources, respectively, while an isolated DC-DC converter is used to provide power to the thruster. To attenuate external disturbance and enhance the robustness of the proposed MVDC system, uncertainty and disturbance estimator (UDE) based controllers are designed for the power electronics converters in the MVDC system.
A far-field radio frequency experimental exposure system for yeast: creating a new standard
(North Dakota State University, 2024) Nevin, Soraya
This paper presents a framework for investigating the potential effects of radiofrequency (RF) exposure on gene expression and cellular processes. Utilizing Schizosaccharomyces Pombe (S. pombe) as a model organism and specialized equipment to create a standard test environment for continued research in this field. Key methods include environmental monitoring, RF antenna configuration, and power density mapping. Results demonstrate the feasibility of the proposed experimental setup. Future directions involve variations in RF frequency and exposure time, as well as exploring underlying molecular mechanisms. This framework aims to lay the groundwork for future experimentation in this field. It emphasizes the need for rigorous scientific inquiry and highlights the potential implications for understanding biological responses to RF exposure. This work provides a roadmap for further investigation into RF effects on gene expression, facilitating advancements in scientific understanding without conducting an actual experiment.
Using wing flap sounds to distinguish individual birds
(North Dakota State University, 2024) Phan, Thinh
Ornithologist often need to recognize individual birds, but traditional invasive methods, such as capturing, marking, releasing, and recapturing of birds, have limitations. To overcome this, researchers use non-invasive alternatives, such as bird vocalizations. In our study, we used wing flap sounds of three male Zebra Finch birds for individual recognition. We achieved identification accuracies ranging from 55% to 100% by using a combination of Principal Component Analysis-K-Nearest Neighbor (PCA-KNN) and Cross-Correlation method on training data and testing data. PCA-KNN allows for dimensionality reduction and pattern recognition, while the Cross-Correlation method bases data analysis on shifting data elements. Our approach can be applied to other bird species and is becoming more accessible due to technological advancements. Non-invasive methods for bird identification are becoming increasingly popular, and our study demonstrates the potential for using wing flap sounds to recognize individual birds.
Dc'sign of Piezoelectric Energy Harvesting and Supercnpacitor Charging Circuits Using Optimized Duty-Cycle of Buck Converter
(North Dakota State University, 2009) Zhao, Zhou
Energy harvesting has been proved to be an innovative solution to replace the batteries in remote power supply applications. Unfortunately, the limited capacity aucl low efficiency of power conversion restrict the practical applications of energy harvesting in daily-life. After a systematic review of research on energy harvesting iu power management perspective, a circuit design, which targets low-frequency mechanical vibrations, is presented. With the classical piezoelectric cantilever setup, the maximum charging current of a supercapacitor can be obtained by optimizing the duty-cycle of a buck converter through software implemented pulse width modulation. The results of experiments prove the capacitive electric model of the piezoelectric element, the existence of maximum charging current of the supercapacitor, and the adaptive control of the designed circuits. With the duty-cycle optimized to 2.17%, a maximum charging current of 17.36 mA is measured, which is approximately 4 times that obtained in related researches. An active radio frequency identification (RFID) application is proposed to utilize the harvested power of 67.2 µW. In addition, the digital design in field programmable gate array (FPGA) is integrated using AMIS 0.5 µrn CMOS process technology to reduce system power dissipation.
Refinementbased Verification of Elastic Pipelined System with Early Evaluation
(North Dakota State University, 2010) Cai, Yangwei
This thesis presents a formal verification procedure to check correctness of the synchronous elastic pipelined system that incorporates early evaluation against its synchronous parent system. Note that the goal of the verification procedure is not to establish the correctness of the algorithm for synthesizing elastic circuits, but instead, to find bugs and formally prove the correctness of elasticized designs with early evaluation. Dataflow through elastic architectures is complicated by the insertion of any number of elastic buffers in any place in the design. Elastic token-flow diagrams are introduced, which are used to track the flow of data in elastic architectures. We provide a method to construct such diagrams. The thesis also develops a highly automated and systematic procedure based on elastic token-flow diagrams that compute functions that map states of elastic systems to states of the synchronous parent systems. Such functions, known as refinement maps, are used to compare behaviors of elastic and synchronous systems and hence prove their equivalence. The effectiveness of this method is demonstrated by verifying 8 synchronous elastic pipelined processor models with early evaluation.
Joint Compression and Encryption
(North Dakota State University, 2010) Ghosh, Ahana
This research work proposes two techniques of joint compression and encryption using Shanon Fano Elias coding and Arithmetic coding respectively. The first scheme proposed is called Adaptive Shannon-Fano-Elias code where it has been observed that the complexity of attacks is exponential in m, where m is the length of the string being compressed. Since mis usually very large (> 220 ), the security of our scheme is very high. The main reason why our scheme's security depends on m is the fact that all attacks require the ciphertext to be scanned from left to right. The algorithm proposed does not compromise in the compression ratio produced by normal Shannon Fano Elia coding. The second scheme proposed uses Arithmetic coding as the compression algorithm as Arithmetic coding is one of the optimal compression techniques that can be used in various applications. The algorithm proposed is proved to be secure under an assumption that the attacker would have access to an algorithm which could decrypt any given message without the knowledge of the key. i
Hybrid Renewable Energy System with Wind Turbine and PV Panels
(North Dakota State University, 2010) Kaderbhai, Munir Shabir
In recent years, there has been increasing concern about the world's heavy dependence on fossil-based fuels and the pollution caused by such fuels. This has sparked an increased interest in the use of other renewable sources of energy. In particular, electricity generation from wind and photovoltaic energy has seen a rapid growth in recent years. One of the most widely used electric machines used in wind energy conversion systems is the doubly-fed induction generator (DFIG) which requires a source and power converter system to feed variable frequency AC voltage to its rotor. Power converter schemes currently used with DFIGs, however, do not address issues such as the absence power factor correction in the rectifier (without complicated control loops), soft-switching in the rectifier, and the possibility of integrating photovoltaic power into the system to provide rotor injection power. This thesis presents a hybrid system in which the rotor power can be drawn either from three-phase AC mains or a photovoltaic (PV) panel-battery combination. While drawing power from AC mains, an AC-DC converter with soft switching and power factor correction is used. While supplying rotor injection power from PV panels, one of the MOSFET switches of the AC-DC converter will be used for boosting the PV panel output voltage. Various stages of the system are implemented and the details of implementation and results are discussed for each stage. Suggestions for future research are also offered.
Verification of Synchronous Elastic Pipelined Systems
(North Dakota State University, 2010) Sarker, Koushik
The constant shrinking of technology has lead to several design challenges that the synchronous design paradigm is unable to cope with. Elastic design is a novel and promising design paradigm that overcomes many of these challenges by using components that are insensitive to the latencies of its inputs. Verification is a critical problem for any design paradigm. The complexity of elastic designs arises when the system is pipelined. We develop formal verification techniques to verify synchronous elastic pipelined systems. Note that the goal of verification is not to establish the correctness of the algorithm for synthesizing elastic circuits, but instead, to find bugs and formally prove the correctness of elasticized designs. We develop two formal verification procedures. The first procedure checks the correctness of elastic pipelined systems against their synchronous parent pipelined systems. The second procedure checks the correctness of elastic pipelined systems against their high-level non-pipelined specifications (such as an instruction set architecture). Datatlow through elastic architectures is complicated by the insertion of any number of elastic buffers in any place in the design. We introduce elastic tokenflow diagrams, which arc used to track the flow of data in elastic architectures. We provide a method to construct such diagrams. We also develop highly automated and systematic procedures based on elastic token-flow diagrams that compute functions that map states of elastic systems to states of their specifications. Such functions, known as refinement maps, are used to compare behaviors of elastic and synchronous systems and hence prove their equivalence. We elasticized a 5-stage DLX processor that enables the insertion of buffers in its data path. We constructed several elastic processors by introducing up to 5 elastic buffers at various places in the data path and verified equivalence with both their synchronous parent pipelined systems and also with their instruction set architecture specifications.
Phasor Measurement Unit Placements for Complete Observability using Linear-Time, Quadratic-Time, and Subquadratic-Time Heuristics
(North Dakota State University, 2010) Saula, Oluwasijibomi
A phasor measurement unit (PMU) is considered to have the potential to improve the efficiency of electric power systems by monitoring, control, and protection. Through measurements of all bus voltages, incoming and outgoing currents, and by subsequent calculation of all phase angles, employing PMUs on every substation in a power system will allow complete observation of a power system. However, having a PMU on every substation may not be economically feasible. Therefore, methodologies must be devised that can monitor a system with the minimum possible number of PMUs. In this paper, we propose six graph theoretical PMU placement heuristics. The proposed heuristics overcome the previous approaches in terms of scalability and execution time. The proposed heuristics are thoroughly compared and benchmarked using standard IEEE bus networks ranging from 14 to 300 buses, and a 2,383 bus system.
A Sensor System Designed for Electrochemical Impedance Spectroscope-based Coating Performance Evaluation
(North Dakota State University, 2010) Shi, Yulong
One of the traditional methods for coating performance evaluation is using Electrochemical Impedance Spectroscopy (EIS). The theory of using EIS to evaluate coating performance is to measure the coating impedance in a frequency range of 0.01 Hz to 100 KHz. Based on the observation of a number of the EIS data, the coating impedance variation trend in low frequency range has a corresponding impedance variation trend in high frequency range. Based on this idea, this work introduces an EIS-based sensor system designed for coating performance evaluation. The sensor system consists of one base station and one sensor node. Wireless communication is realized between the sensor node and the base station. The sensor node is smaller than a business card in size. Evaluation time is less than 10 seconds in the operation frequency range between 20 90 kHz. Three coated panels were used for testing. Testing results are presented for both the EIS instrument and the sensor system. Good correlations indicate that this new technique is valid for EIS based coating performance evaluation in high frequency range. The sensor system is powered by batteries. The power consumption is •120 m W which is lower than the commercial ElS instrument The sensor system is portable and easy to use in real time for fieId coating performance evaluation.
A New Printed Quasi-Landstorfor Antenna
(North Dakota State University, 2011) Al Aziz, Masud
In this thesis, a new type of printed quasi-Landstorfer antenna is present ed. The proposed antenna utilizes parasitic radiators to realize a higher gain. The antc1ma is measured to have a resonant frequency of 2.45 GHz with a return loss of -43 cll3. The gain of the antenna is measured to be 7.2 dBi and can be increased to 8 dBi with the design of a folded slot. The gain increment is achieved without increasing the overall dimension of the antenna. Both the measured and the stimulated radiation patterns indicate that the antenna has a symmetric radiation pattern in the end-fire direction. Moment method based software has been used to stimulate quasi-Landstorfer antenna and the stimulated results are shown to be in good agreement with measured results. The new quasi-Landstorfer antenna presented in this work is 44% smaller in size when compared to the previous Landstorfer antennas and provides higher gain than that of quasi-Yagi antenna. The antenna is well suited for WLAN band applications, such as wireless communications, phased array antenna, and millimeter wave applications.
Impedances of a Cubical Quad Antenna
(North Dakota State University, 1965) Choong, Tsi-lung
Rectangular loop antennas and short electric dipoles are two of the oldest antennas in existence. In 1888, twenty years after Maxwell invented his famous Maxwell's equations, Hertz used these two antennas to prove that high frequency electric energy sources could radiate electro- magnetic waves. The "Cubical Quad" or, simply, "Quad" antenna is a development of the rectangular loop antenna. It consists of a pair of square loops, one-quarter wavelength on a side or one wavelength around the periphery; one loop being driven and the other used as a parasitic reflector. The separation between the two is usually of the order of 0.15 to 0.2 wavelength, with the planes of the loops parallel. While studying the properties of this antenna, it was discovered that little had been done to develop it from a theoretical aspect. The purpose of this thesis is to obtain values of the self and mutual impedances existing in such an antenna array. The values are obtained from mathematical analysis and experimental measurements and may be used in field pattern and gain calculations.
Lithium Ionic Conductivity and Stability Of Cubic Li7La3Zr2O12 Solid Electrolyte A First-Principles Study
(North Dakota State University, 2020) Saha, Sudipto
Garnet structured cubic LLZO crystal (Li56La24Zr16O96) is one of the most promising solid electrolytes for next-generation solid-state lithium-ion batteries. Ab initio molecular dynamics simulations have been employed to study the impacts of lithium vacancy defect and doping concentration on the lithium ionic conductivity and stability of LLZO. The number of lithium atoms in a unit cell of LLZO has been reduced from 56 to 53, where 56 lithium atoms represent the structure of stoichiometric LLZO, i.e., Li7La3Zr2O12. Similarly, the effect of Al and Ga doping on the conductivity and stability of LLZO material was also investigated. Our computational results confirm that both the defects help in enhancing the conductivity of LLZO and the concentration of defect introduced controls the trade-off between the conductivity and stability. Overall, this study provides a valuable insight into the enhancement of conductivity of cubic LLZO garnet material along with structural stability.
Using Classification and Clustering to Predict and Understand Student Behavior in an Innovation-Based Learning Course
(North Dakota State University, 2020) Singelmann, Lauren Nichole
One of the Grand Challenges for Engineering is advancing personalized learning, but challenges remain to identify and understand potential student pathways. This is especially difficult in complex, open-ended learning environments such as innovation-based learning courses. Student data from an iteration of an innovation-based learning course were analyzed using two educational data mining techniques: classification and clustering. Classification was used to predict student success in the course by creating a model that was both interpretable and robust (accuracy over 0.8 and ROC AUC of over 0.95). Clustering grouped student behavior into four main categories: Innovators, Learners, Surveyors, and Surface Level. Furthermore, noteworthy variables from each model were extracted to discover what factors were most likely to lead to course success. The work presented contributes to gaining a better understanding of how engineering students innovate and brings us closer to solving the Grand Challenge of advancing personalized learning.
Extraction of Electromagnetic Properties of Metamaterials with Branch Compensation from Phase Tracking
(North Dakota State University, 2020) Lewis, Jacob Christian
In the field of electromagnetism, there are materials known as metamaterials which exhibit unique properties that can be exploited. Permittivity, defined as capacitance per meter, of a metamaterial can vary over frequency , time, or even be negative. This can be useful for tuning antennas, changing their operating frequency or direction of propagation, or even designing cloaking systems. However, the theory behind metamaterials needs to be studied further. One of the biggest issues to address is in determining the constitutive parameters of metamaterials which may be varying. Previous research has shown the issue of branches, or mathematical discontinuities, occurring in the derivation of permittivity from the scattering parameters of a metamaterial. This thesis provides further understanding to the theory behind these branches and presents a new method to compensate for them. This new method, called the phase tracking method, may be considered a modern adaptation of the Nicolson-Ross-Weir method.
Novel K2W7O22/Ti3C2 Nanocomposite-Based Sensor Device for Breath Acetone Analysis in Diabetic Patients
(North Dakota State University, 2020) Ama, Obinna Henry
Acetone in exhaled breath is gaining attention as a non-invasive means of quantifying blood glucose levels in Diabetics. This calls for development of novel biosensors for the detection of trace concentrations of acetone present in human breath. Traditional gas detection systems, such as GC/MS and chemiresistive sensors, are currently used for this purpose. However, these systems have limitations with regards to size, cost, and operating temperature. This work presents the K2W7O22/Ti3C2 nanocomposite sensor as breath acetone sensor that overcomes the limitations in traditional detection systems. Sensing experiments were conducted using 5 different sensor materials in varying ratios. KWO/Ti3C2 - ratio 2:1 (annealed) and KWO/Ti3C2 - ratio 2:1 (Unannealed) showed excellent sensitivity to 2.85ppm and 5.4ppm acetone concentration. These materials were then implemented in a prototype device. Material and device test results confirm the potentials of the novel KWO/Ti3C2 nanocomposite as a good sensor for breath acetone detection.
A Negative Dielectrophoresis Based Method of Detecting Pancreatic Cancer Antigen CA 242 in Serum
(North Dakota State University, 2020) Afrose, Sharmin
Patients with pancreatic cancer in metastasis rarely survive, thus the need for diagnostic tools for early stage detection. Current techniques such as ELISA and SPR are complex and expensive and cannot detect cancer in its early stages. Cancer Antigen 242 (CA 242) is a potential protein biomarker of pancreatic cancer with high sensitivity and specificity. This thesis presents a negative Dielectrophoresis (DEP) based method of detecting pancreatic cancer protein biomarker CA 242 in serum. A spectrum of concentration levels was generated with a cut off level 20 U/mL using a transduction mechanism with negative DEP spectroscopy, light scattering, and image processing. This was a fast and cost-effective method to diagnose early stage pancreatic cancer. This thesis also presents the design and simulation of an electrode modified to increase the electric field gradient with reduced heat generation and a concentration prediction model to predict concentrations from the generated spectrum of experiments.
State-Space Randles Cell Model for Instrument Calibration
(North Dakota State University, 2020) Fonseca, Aaron James
It is desirable to calibrate electrochemical impedance spectroscopy (EIS) instrumentation using a Randles circuit. This presents a challenge as realistic loads, simulated by this circuit, contain theoretical components (Warburg elements) that are difficult to model. This thesis proposes a state-space solution to this problem and explores the process of realizing a digital high-accuracy approximation of a Randles circuit for the purposes of verifying and calibrating EIS instrumentation. Using Valsa, Dvo{\v r}{\'a}k, and Friedl's network approximation of a Warburg element, a collection of state-space relations describing the impedance of a Randles circuit are derived. From these equations the process of realizing a digital system is explored; this includes a discussion on methods of discretization, an overview of the challenges of realizing digital filters, and an analysis of the effects that finite word-length has on the accuracy of the model when using fixed-point hardware.

Browse

Recent Submissions