Graduate School of Engineering and Science
Permanent URI for this collectionhttps://hdl.handle.net/11693/115678
Browse
Recent Submissions
Item Open Access Optimized RF safety monitoring for cerebellar magnetic resonance imaging at 7T(2024-09) Mahmoudalilou, Elnaz MahmoudiThe primary objective of this thesis is to develop optimized RF safety assessment techniques for cerebellar imaging using 7T MRI systems, with a specific focus on Spinocerebellar Ataxias (SCAs). Due to the unavailability of detailed electro- magnetic simulation data from the manufacturer, this study focused on predicting the electromagnetic behavior of the Nova 8Tx/32Rx coil’S 8 pTx channels to en- sure accurate and safe imaging. Accurate prediction of the coil’s electromagnetic performance is essential for both RF safety and imaging quality, particularly in managing RF exposure and minimizing tissue heating. The approach involved replicating the electromagnetic fields of the Nova coil through simulations, validating these predictions against experimental measure- ments, and implementing an algorithm for calculating the temperature-based Virtual Observation Points (tVOPs) in future works for rapid RF safety assess- ments. While the initial simulations captured key aspects of the coil’s B1+ field distribution, discrepancies between predicted and experimental results revealed challenges, especially in random-phase shimming configurations. The limitations of using ideal current sources and the reduced dataset for optimization highlighted the need for more comprehensive data and realistic models. The findings underscore the importance of integrating empirical measurements with refined simulations to bridge the gap between theoretical models and real- world performance. Future work should focus on enhancing current source mod- els, mitigating noise and experimental inaccuracies, and expanding the applica- tion of these techniques to broader clinical scenarios. By addressing these chal- lenges, this research can contribute to improving both the safety and quality of high-field MRI, ultimately advancing its reliability for both clinical and research applications.Item Embargo Reconfigurable CNN accelerator design using dataflow analysis(2024-09) Kalay, AlperenDataflow reconfigurability plays a crucial role in Convolutional Neural Network (CNN) acceleration by determining the optimal dataflow pattern for convolution operations. Fully reconfigurable architectures provide versatility and high resource utilization by supporting multiple dataflow options, but this comes with increased design complexity and operational overhead. On the other hand, non-reconfigurable architectures, optimized for a single dataflow pattern, deliver high efficiency for specific tasks but lack adaptability. This thesis introduces a novel intermediate dataflow reconfigurable CNN accelerator that balances flexibility and efficiency by integrating key dataflow patterns, enhancing adaptability and performance across diverse CNN applications. Through a detailed analysis, key dataflows are identified, and a unique architectural unit is developed for dataflow selection, with an average of 0.15% excess latency compared to the optimal scenario. Our specialized systolic array architecture accommodates various kernel sizes, providing an additional layer of reconfigurability. Our architecture requires 39% less area and 35% less power than fully reconfigurable designs. Additionally, it delivers an average of 33% better performance compared to non-reconfigurable architectures. In terms of efficiency, it provides a 7% increase over fully reconfigurable designs and outperforms non-reconfigurable options by up to 3.57X.Item Embargo Real time identification of cornering coefficients and ideal twin driving assistance(2024-08) Keleş, Ahmet FarukCornering coefficients play a crucial role in vehicles’ lateral and longitudinal dynamics. They depend on many factors, including environmental factors such as road conditions. In this study, a method that identifies cornering coefficients in real-time by utilizing deep neural networks is developed. Three different neural network architectures with two different datasets are compared for this identification. Results show that a fully connected network trained with time-varying cornering coefficients performs best. Compared to constant cornering coefficients, this method improves the lateral force estimation between 42.62-75.47 % in experiments conducted on a 1/8 scale four-wheel drive four-wheel steering vehicle. A control method that utilizes the identified cornering coefficients to cancel out the changes in cornering coefficient by utilizing 4 wheel drive 4 wheel steering system is developed. The control method utilizes a nonlinear model predictive controller. The control system uses the driver’s control inputs in an ideal front wheel drive front wheel steering twin of the vehicle with constant cornering coefficients set by the driver, to obtain reference velocities. The nonlinear model predictive controller can calculate the optimal control inputs for a 4-wheel drive, 4-wheel steering vehicle from these references. The simulation results show that this control method improves reference tracking by 85 % compared to a conventional configuration, i.e., a front wheel drive front wheel steering vehicle. The controller is tested on an experimental setting and found to be improving the results by 12.12 %. The reduction in improvement can be attributed to noise in measurements and delays in the control system.Item Open Access Investigating hyperlipidemia-driven organelle stress and neuroinflammation on the mouse cerebral cortex: insights into the intervention of perk pathway(2024-09) Kızıldağ, FulyaDeficits in the metabolism of lipids called hyperlipidemia have been linked to a higher risk of developing neurodegenerative diseases. Protein Kinase RNA-like Endoplasmic Reticulum Kinase (PERK) signaling is crucial in cellular homeostasis. Abnormalities in the PERK have been associated with neurodegeneration. Mitophagy and the PERK pathway emphasize how cellular stress responses are regulated to preserve cellular homeostasis and mitochondrial quality control. The activity of main mitophagy regulators, such as Parkin and PINK1 (PTEN-induced kinase 1), is regulated by the phosphorylation of eukaryotic initiation factor 2 alpha (eIF2α) by PERK. If lipid metabolism is at a high level, abnormalities in the mitochondria and endoplasmic stress (ER) emerge. During the ER stress activation, the PERK pathway is induced, and mitophagy is blocked, causing an enhancement in the neuroinflammation. The underlying molecular mechanism by which hyperlipidemia impacts the PERK pathway and mitophagy in the cerebral cortex, as well as the relationship between mitophagy and neuroinflammation, is not fully understood. In this study, Apoe-/- and C57BL/6 mice were given a chow or western diet to stimulate hyperlipidemia. Moreover, western diet-fed Apoe-/- mice were injected with PERK inhibitors, GSK2606414 and Trans-ISRIB, intraperitoneally for six weeks to suppress the PERK pathway. This study explores the effects of hyperlipidemia on the PERK pathway, inflammatory and mitophagy markers in the cerebral cortex of chow and western diet-fed C57BL/6 and Apoe-/- mice and investigates whether the inhibition of the PERK pathway can change the levels of inflammatory and mitochondrial markers in the cerebral cortex of hyperlipidemic mice subjects. mRNA and protein expression levels of mitophagy and inflammatory markers were assessed using the RT-qPCR and western blot, respectively. PERK pathway activation under hyperlipidemia conditions was not determined. Nevertheless, significant alterations in mitophagy markers and inflammation were detected in Apoe-/- mice apart from the diet. Furthermore, significant alterations were not seen in the PERK pathway markers; however, mitophagy was stimulated, and some inflammation markers were significantly decreased mildly at the cortical tissue of WD-fed Apoe-/- mice administrated with PERK pathway inhibitors, GSK2606414 and Trans-ISRIB. Besides, no statistically significant changes were observed in the transcript levels of the inflammatory markers. Taken together, hyperlipidemia did not cause the PERK pathway to be activated in the cerebral cortex of mice; nevertheless, it mildly altered inflammation and caused mild effects of the dysregulation of the mitochondria by hyperlipidemia independent from the PERK pathway. Furthermore, although the PERK pathway was not inhibited by the administration of PERK pathway inhibitors, mitophagy was induced, and inflammation was decreased mildly. Targeting the PERK pathway with GSK2606414 and Trans-ISRIB inhibitors from the cerebral cortex would not be a therapeutic approach for neurodegenerative diseases.Item Open Access Impacts of high-fat diet and genotype on blood-brain barrier and synaptic integrity in mouse cerebral cortex: an exploration of perk pathway(2024-09) Şeker, BüşranurHigh-fat diet intake can induce hyperlipidemia and result in cognitive decline by causing endoplasmic reticulum stress, decreased blood-brain barrier, and synaptic integrity. The protein kinase R-like endoplasmic reticulum kinase (PERK) pathway is one of the arms of the unfolded protein response, which is activated by endoplasmic reticulum stress. The PERK inhibits the global protein translation while allowing the translation of certain proteins that are involved in inflammation and apoptosis. Due to its apoptotic properties, it is thought that the PERK pathway causes neurodegeneration. To study the effects of hyperlipidemia, a high-fat diet-fed Apoe knock-out mice model (Apoe-/-) is appropriate. Knocking out the Apoe in mice makes the animal model more prone to high-fat diet-induced hyperlipidemia. In the cerebral cortex of these animals, endoplasmic reticulum stress, blood-brain barrier, and synaptic integrity markers are checked at protein and mRNA levels. No changes are observed in the PERK pathway markers besides phosphorylated eukaryotic Initiation Factor 2. Additionally, there is a significant increase in blood-brain barrier marker Claudin-5 levels in Apoe-/- mice fed with a high-fat diet. There is also no significant change in synaptic integrity markers. In the second part, the effects of the PERK pathway inhibition are checked with integrated stress response inhibitor and GSK2606414 in the high-fat diet-fed Apoe-/- mice cerebral cortex. There are no significant alterations in BBB and synaptic integrity when the animals are injected with inhibitors. In conclusion, this study investigates the effects of high-fat diet induced hyperlipidemia in the cerebral cortex of Apoe-/- mice on ER stress, blood-brain barrier, and synaptic integrity. In the cerebral cortex region, the PERK pathway-related ER stress is not observed, and synaptic integrity remained unchanged while the blood-brain barrier is affected. Moreover, the effects of the PERK pathway inhibition are researched, and there is no inhibition effect observed in the cerebral cortex region.Item Open Access Novel deep learning approaches for functional FMRI data analysis(2024-08) Bedel, Hasan AtakanFunctional MRI (fMRI) has revolutionized our ability to analyze brain activity by providing insights into high-dimensional, time-series data. Despite advancements, existing methods often fall short in their ability to effectively capture contextual representations across varying time scales and interpret the resulting data. Addressing these challenges, this thesis introduces three innovative ap-proaches: BolT, GraphCorr, and DreaMR, each designed to enhance the analysis and interpretability of fMRI data. BolT represents a significant advancement in modeling fMRI time series by utilizing a blood-oxygen-level-dependent trans-former architecture. This model incorporates a novel fused window attention mechanism, which enables the extraction of both local and global representations by processing temporally-overlapped windows and employing cross-window regularization. BolT’s approach improves upon existing methods, offering enhanced sensitivity and aligning with key neuroscientific findings through extensive experimentation. GraphCorr addresses limitations in static functional connectivity (FC) features used in classification models by introducing a graph neural network-based plug-in. This method captures dynamic latent FC features while preserving dimensional efficiency, employing a node embedder and lag filter module to re-fine temporal information. The integration of these features through a message passing algorithm significantly enhances the performance of baseline classification models, as demonstrated through comprehensive testing on public datasets. Finally, DreaMR tackles the interpretability of deep fMRI classifiers through a novel diffusion-driven counterfactual approach. By using fractional multi-phase-distilled diffusion, DreaMR generates high-fidelity counterfactual samples and employs a transformer architecture to account for long-range spatiotemporal contexts. This method surpasses traditional counterfactual techniques in both fidelity and efficiency, offering a more precise and actionable explanation of classifier decisions. Together, these contributions advance the field of fMRI analysis by improving model performance and interpretability, thereby facilitating more effective and insightful neuroimaging research.Item Embargo Development and implementation of a concurrent path following and body motion control system using a scale prototype vehicle(2024-09) Dağ, DoğaThe automotive industry’s future lies in developing autonomous vehicles, which rely on three fundamental components: perception, planning, and actuation. The actuation component, in particular, requires a robust and reliable control system to ensure the planned trajectory is executed with high fidelity. Furthermore, ad- vancements in active suspension technology, driven by cost reductions, have en- abled the individual adjustment and actuation of suspension components, thereby enhancing passenger comfort. A prototype vehicle is essential to validate these systems experimentally. To balance cost efficiency, ease of use and maintenance, and minimize risks to the en- vironment, a 1/8-scale miniature vehicle was developed. This scaled-down model incorporates all the necessary and existing systems of a full-sized vehicle, like all-wheel drive, four-wheel independent steering, disc brakes, and active suspen- sion, allowing for a wide range of system tests, regardless of their power demands, facilitated by wireless communication via the Robot Operating System (ROS). For control purposes, a Model Predictive Control (MPC) system was designed for path following, while a Linear Quadratic Regulator (LQR) was implemented for body motion control. In simulations, the MPC controller achieved a maxi- mum deviation of 0.3 meters from the planned path, while real-life tests on the prototype vehicle resulted in a deviation of 0.51 meters. During simulations, the active suspension system demonstrated a 48.38% improvement in pitch perfor- mance and a 50.92% improvement in roll performance. Real-world tests showed a 67. 54% improvement in pitch and a 2.44% improvement in roll performance. However, the performance of the MPC controller was adversely affected by system delays, as the controller operated on an external PC during both path-following experiments. This research focuses not only on controller design for vehicles but also tests this system on a scale of real vehicles to assess their performance and ensure that these systems can work on real passenger cars.Item Embargo Investigations into the evolution of heated liquid films(2024-08) Mohamed, Omair A. A.The evolution of the free surface of a heated liquid film is directly tied to the performance and efficiency of various industrial systems. Therefore, we investigate the spatiotemporal evolution of heated liquid films across a range of different settings by formulating distinct of hydro-thermal models taking into account the effects of inertia, thermocapillarity, evaporation, gas shear, and thermal radiation, where we direct our modeling efforts in each problem on the most dominant physical phenomena. In liquid flows characterized by relatively low Reynolds numbers belonging to the drag-gravity flow regime, we model the hydrodynamics of the film using the long-wave expansion (LWE) methodology and perform linear stability analyses focused on the thermocapillary and evaporative instabilities, as they have a primary influence on the film’s evolution in this flow regime. Consequently, the evaporation process is governed by the competition between thermodynamic disequilibrium and diffusion effects dependent on the interface’s curvature. We modify the kinetic-diffusion evaporation model of Sultan et al. [Sultan et al., J. Fluid Mech. 543, 183, (2005)] and combine it with long-wave theory to derive a governing equation encapsulating the coupled dynamics. We then utilize linear stability theory to derive the system’s dispersion relationship, in which the Marangoni effect has two components. The first results from surface tension gradients driven by the uneven heat flux, while the second arises from surface tension gradients caused by imbalances in vapor diffusion. These two components interact with evaporative mass loss and vapor recoil in a rich and complex manner. Moreover, we identify an evaporation regime where a volatile film is devoid of evaporation instabilities. Furthermore, we investigate the effect of film thinning on its stability at the two opposing limits of the evaporation regime, where we find its impact in the diffusion-limited regime to be dependent on the intensity of evaporative phenomena. Finally, we conduct a spatiotemporal analysis which indicates that vapor diffusion effects are correlated with a shift towards absolute instability. In the second problem, we study the spatiotemporal evolution of an evaporating liquid film sheared by a gas and consider both the inertial and thermal instability modes, where the shearing gas is modeled by imposing a constant shear stress along the liquid’s interface. Interestingly, it’s inclusion in the problem allows the utilization of a one-sided evaporation model, which is precisely the transfer-rate-limited case of the first system we investigated. Once more long-wave theory is used to derive the an evolution for the liquid film which incorporates the role of the shearing gas. Afterwards, linear stability theory is used to investigate the temporal and spatiotemporal characteristics of the flow, where it is found that the evaporation of the film promotes absolute instabilities and can cause convective/absolute transitions. We also find that counter-flowing shearing gas can suppress the inertial instability affirming similar conclusions found by previous studies for a strongly confined isothermal film. Furthermore, the evolution interface equation was solved numerically to explore the film’s nonlinear stability. Moreover, we employ self-similarity analysis to probe the shear stress’s effect on the film’s rupture mechanics. In the third problem we research, the liquid flow’s Reynolds number is relatively high, and hence we utilize the weighted-residual integral boundary layer (WIBL) technique [C. Ruyer-Quil and P. Manneville,” Eur. Phys. J. B, 15, 357, (2000)], and direct our attention at directly simulating the temperature field across the film using reduced models. The WIBL hydrodynamic equations are derived expressions obtained via the boundary layer approximation, while the thermal profile is modeled by employing an asymptotic expansion which produces a hierarchy of models in which enhanced sophistication is offset by higher complexity and computational cost. These models are solved numerically revealing how the temperature field across the film is governed by a balance between the conduction across both the liquid film and the solid surface, and their respecitve radiative emissions, wherein these two transfer phenomena are linked through two corresponding dimensionless numbers associated with both the liquid film and the solid surface.Item Embargo Homogenization-based computational design and two-scale performance optimization of electroactive structures(2024-09) Dedeoğlu, BerkanMacroscopic materials and structures with enhanced characteristics have been extensively studied in the context of solid mechanics. The advantages of mi- crostructured materials with active constituents have been reported in the lit- erature. In view of tunable microstructures, further intriguing traits of such materials can be achieved through imposing external stimuli. The ultimate goal of the present study is to establish a homogenization-based computational design framework to actively control the time-varying macroscopic stress response and behavior of structures with piezoelectric constituents. This is accomplished by temporally adapting the macroscopic electric field enforced on a microstructure and controlling the time-variation of the macroscopic electric potential imposed on a macroscopic solid. This periodic microstructure is optimized in a non- restrictive design space that embodies not only the topology, but also anisotropic material orientation and the unit cell geometry. In order to enrich the optimiza- tion space to capture the intriguing time-varying mechanical aspects, additional optimization variables, namely performance variables, are developed. Extensive numerical investigations are conducted to test the limits of this framework based on the discreteness of the microstructure and the accurateness of attaining the targeted mechanical behavior. The overall computational work is implemented through a parallel C++-based in-house FE program.Item Open Access Analysis of speech content and voice for deceit detection(2024-09) Eskin, Maria RalucaDeceptive behavior is part of daily life, often without being recognized, leading to severe repercussions. With the recent improvements in machine learning, more reliable detection of deceit appears to be possible. Although current visual and multimodal models can identify deception with adequate precision, the individual use of speech content or voice still performs poorly. Therefore, we systematically analyze such essential communication forms focusing on feature extraction and optimization for deceit detection. To this end, we assess the reliability of employing transformers, spatial and temporal architectures, state-of-the-art pre-trained models, and handcrafted representations to detect deceit patterns. Furthermore, we conduct a thorough analysis to comprehend the distinct properties and discriminative power of the evaluated methods. The results demonstrate that speech content (transcribed text) provides more information than vocal characteristics. In addition, transformer architectures are found to be effective in representation learning and modeling, providing insights into optimal model configurations for deceit detection.Item Embargo Capturing the dynamic scaffold properties of hybrid GelMA based microgels toward tissue engineering and organ-on-chips(2024-09) Çınar, Aslı GizemMicrogels have emerged as versatile materials in tissue engineering, drug delivery, and organ-on-chip (OoC) platforms due to their small scale, uniformity, and customizable properties. Their adaptability as injectable materials and dynamic scaffolds makes them promising candidates for a wide range of biomedical applications. However, traditional methods for characterizing their physical and mechanical behaviors, designed for bulk hydrogels, do not capture the unique properties of microgels, which differ significantly in terms of size and surface-to-volume ratio. This work explores the physical properties of Gelatin Methacryloyl (GelMA)-based Collagen and Hyaluronic Acid Methacrylate (HAMA) hybrid microgels produced via droplet microfluidics, employing novel assays tailored specifically to their micro-scale. Real-time observation of their swelling and degradation properties is carried out using a custom-made platform enabling the tracking of individual microgels, and electron microscopy provides insights into their internal structures, revealing previously unobserved behaviors. We have shown the interpenetrating network formation when GelMA and Collagen are used; and copolymer formation when GelMA and HAMA are used. Under the effect of Collagenase and Hyaluronidase, the individual microgels showed different degradation mechanisms, which have proven to be affected by crosslink densities, enzyme-substrate specificity, enzyme saturation, and properties of the individual network components. The work is extended by focusing more on the temporal profiling of GelMA and HAMA hybrid microgels' behaviors under enzymatic degradation, examining how volume, mechanical properties, and surface features evolve over time, simulating the dynamic conditions encountered in vivo during especially tissue engineering applications. We found that instead of carrying out separate assays to understand the changes, a more holistic approach to evaluating the aforementioned properties gives a more thorough discussion. This approach revealed that changing the ratios of GelMA against HAMA affects the crosslink densities, network formation, and ultimately degrative behaviors. We have observed, for the first time in droplet microfluidics, that a certain combination of GelMA HAMA results in microgels with a network gradient, getting denser towards the center, while the other combinations only increased the crosslink densities without altering the porous homogeneity. Furthermore, the number of microgels exposed to the same concentration of enzyme is altered to emulate different injection volumes into similar tissues, or the enzyme concentration is altered to emulate injection into different tissues. These assays showed the sensitivity of degradation profiles against enzyme saturation and competition. Meanwhile, the stiffness and surface morphology changes of microgels during degradation are examined, revealing the importance of network homogeneity in presenting stable mechanical properties during degradation. Lastly, drug release from these scaffolds is modeled for prospective applications, and their relation to scaffold properties is evaluated. Overall, this thesis is poised to discover the peculiar behaviors of GelMA hybrid microgels produced with droplet microfluidics uncovering the importance of carrying out investigations true to the sample at hand and the conditions that will be imposed upon them during application.Item Open Access Genetically designed microbes for bioimaging and biosensing(2024-09) Yavuz, MerveThe advantageous approach to the utilization of the microbes for bioimaging and biosensing underlies under their active motility and self-propulsion characteristics besides their easy bioengineering feature to gain multi-functional activities. The emerging developments make use of microorganisms as therapeutic agents in disease diagnosis and treatment. The dynamic nature of the habitat forces the microorganisms to acclimate themselves to changing living conditions via evolving exclusive bio-functionalities for their survival. Therefore, the living microorganisms producing functional materials serve as a biohybrid system with unprecedented potential for enhancing the detection of a disease biomarker molecule or meeting the great need in cancer diagnosis. The synthetic biology approach, a multidisciplinary field of science, gives the ability to engineer and modulate the microorganisms to redesign existing natural pathways, resulting in the gain of the desired function. Inspiring form nature, the biomineralization of iron-oxide materials is demanding for their potential usage in antitumor effect due to their easy modulation, stability, and magnetic properties. Furthermore, the certain respiratory capacities of electrochemically active microbes enable the respiration of diverse inorganic and organic molecules for their survival in redox-stratified environments. The ability of exchanging electrons with electrodes possesses several diverse biotechnological applications like the construction of microbial fuel cells, electro-fermentation, and electro-genetics. In this thesis, the microbes were engineered for their utilization in bioimaging and biosensing applications. Firstly, intracellular and extracellular magnetite accumulating Escherichia coli bacterial cell machineries were constructed as contrast agents for the MRI scanning, promising for a cancer diagnostic. Secondly, the intracellular magnetite accumulating bacterial cells, possessing all the redox reactions that readily take place in their cytoplasm via synthetically produced proteins, were further engineered to improve their targeting capability for breast cancer tumor cells by displaying a certain nanobody on the cell surface. Thirdly, electronic sentinel bacterial cells were designed utilizing the electron transfer modules for extracellular electron consumption by targeted acceptors for their wireless biomonitoring applications upon detecting a disease molecule. The methodologies described in this thesis are envisioned as promising tools for diagnostic applications.Item Embargo Unraveling structure-functionality relationships of shape-defined Cu2O nanocrystal model catalysts for methanol decomposition(2024-08) Karaca, KaanMethanol is one of the centerpieces of the chemical industry as a C1 building block and an intermediate producing high-value chemicals such as formaldehyde, methyl methacrylate, methyl tertiary-butyl ether/ tertiary-amylmethylether, and acetic acid. The global demand for methanol is expected to grow exponentially due to its applications in hydrogen production, direct methanol fuel cells, and olefin production via the Methanol to Olefins (MTO) processes. Cu-based catalysts have been widely studied both in academia and in the industry to transform methanol into value-added chemicals at the industrial scale. Some of these academic fundamental research studies have been performed either under ultra-high vacuum (UHV), cryogenic temperature conditions utilizing single-crystal nanocatalysts or under industrially relevant high temperature-pressure conditions utilizing complex mesoporous catalysts resulting in complex data which is challenging to analyze in a conclusive manner to obtain reliable mechanistic information due to the presence of the well-known limitations in heterogenous catalysis called “the materials gap” and “ the pressure gap”. Thus, uniquely defined model catalysts are required to bridge these gaps by offering well-ordered surfaces that can be studied under ambient conditions. This thesis focuses on the structure-functionality relationships of shape-defined Cu2O model catalysts for methanol decomposition. Cubic and octahedral Cu2O nanocrystal catalysts were synthesized and characterized by various ex-situ methods such as Scanning Electron Microscopy (SEM), X-Ray Diffraction (XRD), X-Ray Absorption Near Edge (XANES), Extended X-Ray Absorption Fine Structure (EXAFS), Attenuated Total Reflectance Infrared Spectroscopy (ATR-IR), X-Ray Photoelectron Spectroscopy (XPS) and H2-Temperature Programmed Reduction (H2-TPR). The nature of the surface-active sites were characterized by CO adsorption via in-situ Fourier Transform Infrared Spectroscopy (in-situ FTIR) and the morphology-dependent methanol and formaldehyde decomposition properties were studied via in-situ FTIR and Temperature Programmed Desorption (TPD). The results showed that c-Cu2O and o-Cu2O have distinct structure-functionality relationships for methanol decomposition. It is proposed that the labile surface oxygens that can be readily donated from the c-Cu2O surface can facilitate low-temperature (T ≤ 250 °C) methanol/methoxy oxidation to formates which in turn yield predominantly CO2 and H2O as the total oxidation products. In contrast, limited reducibility of the c-Cu2O surface only allows methanol/methoxy oxidation to first formaldehyde and then to dioxymethylene, eventually yielding predominantly CO and H2 as the thermal decomposition products, indicating the predominance of dehydrogenation catalytic pathways rather than total oxidation thus, unraveling the structure-functionality relationships of shape-defined Cu2O nanocrystal model catalysts for methanol decomposition.Item Open Access The role of extracellular vesicles in pediatric COVID-19 patients with asthma(2024-09) Çetinkaya, Pınar GürSARS-CoV-2, the causative agent of coronavirus disease 2019 (COVID-19), has infected millions of people, and asthma was initially considered a risk factor for COVID-19. Although numerous studies on asthma and COVID-19 have been conducted, the role of plasma-derived extracellular vesicles (EVs) in COVID-19 patients with asthma remains unknown. In this study, we assessed the influence of EVs from healthy controls, severe and mild COVID-19 pediatric patients with or without asthma during acute and convalescent periods on healthy naïve CD4+T cells and monocytes. While plasma cytokines and anti-SARS-CoV-2 antibodies were similar between the groups with and without asthma, immune responses varied depending on the severity of COVID-19. In the severe acute group, whereas all cytokines increased, IFNγ, CD4+T cell counts, and monocyte numbers decreased. Stimulating healthy cells with EVs from severe acute patients led to increased PDL1 expression, Th2 and Treg cell proportions, decreased IFNγ secretion, Th1, and Th17 cell ratios. Patient EVs also reduced proinflammatory cytokine secretion from monocytes. Severe acute patient EVs caused a decline in healthy CD4+T cell and monocyte populations. Overall, our results indicate immunological responses and EV-related outcomes depending on the severity of COVID-19 rather than the presence of asthma, immunosuppression seen in severe acute COVID-19 and potential contribution of EVs to this immunosuppressive pattern in severe cases.Item Embargo Process development for microfabrication of phase reversal CMUT devices for structural health monitoring and development of dynamic characterization processes for MEMS applications(2024-08) Küçük, Merve MintaşIf appropriately designed, Capacitive Micromachined Ultrasonic Transducers (CMUTs) offer advantageous properties such as low cost, small size, low impedance, and environmental friendliness, over piezoelectric transducers. These advantageous properties of CMUTs enable the CMUT devices to be employed in a large area of applications, such as medical applications and non-destructive testing (NDT) applications. CMUT devices and technologies that are heavily developed for medical applications also shed light on the development of CMUT devices to be used in Structural Health Monitoring (SHM) applications for civil infrastructures. Continuous monitoring of the signals produced by the sudden changes happening within civil infrastructures such as bridges or railways may give crucial information about the health of these structures. The rapid release of localized strain energy, which generates Acoustic Emission (AE) waves, is an important indicator of the state of the health of a structure. Detecting AE wave signals may give significant clues about damage formation such as impact, crack initiation, or crack growth. Because AE waves are scattered among a broad range of frequencies, sensing of such AE waves should also be done in broadband, and sensors are preferred to be highly sensitive among such band. For real-life applicable developments, it should be also considered that the environment of the real-life application may be very noisy due to many unrelated reasons, which makes employment of the CMUTs developed in a tightly controlled laboratory environment unpractical for the real-life applications. The noise may often be induced by the noise interferences that are produced by a variety of events that are not needed to be detected. To prevent misjudgments, it is important to differentiate between noise interferences and relevant AE signals, as the presence of significant noise can hinder the detectability of AE waves associated with structural damage. In this process development for CMUT prototype microfabrication study, we collaborated with a group of researchers who have introduced a new approach to designing broadband CMUTs, as well as a unique type of CMUT combination that uses phase-reversal (PR) of generated electrical current for detecting a wide range of mechanical vibration wave frequencies and reducing unwanted noise. By considering the simplest combination of two CMUT cells, the theoretical study, supported by FEM simulations, demonstrated that reversing the electrical current phase of one cell can create low-frequency and high-frequency stopbands for noise rejection, which is applicable for CMUTs operating in air damping. The primary objective of this thesis study is to develop microfabrication processes to microfabricate PR-CMUT devices to bridge the gap between theoretical design and real-world application of PR-CMUT devices. These PR-CMUT arrays that are designed for wafer-scale batch-compatible manufacturability have a flat passband in the 200-250 kHz and 200-300 kHz frequency ranges and two improved stopbands on both sides of the relevant frequency ranges. The photolithography masks, compatible material selections, and microfabrication process flows (integration processes) required for the microfabrication of these PR-CMUT devices were designed considering the capabilities of our cleanroom facility. Microfabrication of the devices was tried multiple times, and in line with the problems encountered in these processes, the microfabrication process flows were updated and the PR-CMUT devices were tried to be produced in multiple iterations. Unit processes, and multiple integration processes were developed and completed. Possible solutions to be implemented in the future microfabrication studies were determined. Additionally, dynamic characterization of individual circular geometry CMUT membranes were explored using a ZYGO Optical Profilometer. With this measurement tool (ZYGO), it is possible to measure CMUT device membrane displacements precisely when the membrane of the CMUT device is moving (vibrating) dynamically. Results obtained from ZYGO Optical Profilometer tool were compared with the impedance analyzer results. It was shown that the resonance frequency of a circular membrane CMUT device can be observed with the ZYGO Optical Profilometer. Furthermore, based on the conclusions from the studies in this thesis, future studies are suggested for further development towards realization and characterization of these PR-CMUT MEMS (MicroElectroMechanical System) devices.Item Embargo Finding expert developers using artifact traceability graphs(2024-09) Hanhan, İdilMentoring is a commonly used practice in the software industry where mentors and mentees are matched to ease the onboarding process of the mentee, who is a newcomer. Also, during a project’s life cycle, developers work on sections of the codebase that are unfamiliar to them. Both cases raise the task of finding an expert developer to contact for possible questions. With this study, we aim to construct an algorithm that recommends expert developers for a specific part of the codebase, namely folders, files, and methods, based on previous developer activities such as commits and code reviews. We construct an artifact traceability graph using commit history, method change history, code review history, and issue history. The relationships in the graph are weighted according to recency and a weight coefficient we determine intuitively. Utilizing this graph, we calculate a score representing the developer’s expertise level on a folder, file, or method, and recommend developers with the highest expertise. To evaluate the success of our algorithm, Expert Developer Finder, we compare its recommendation with the developers who commented on related issues. We run our algorithm on three open-source projects - Nutch, OpenNLP, and Curator. On average, for weighted recommendations, we reached up to 84% accuracy for folders, 82% accuracy for files, and 88% accuracy for methods. On average, for unweighted recommendations, we reached up to 84% accuracy for folders, 84% accuracy for files, and 93% accuracy for methods. We believe that our results show that the Expert Developer Finder algorithm is able to recommend experts by utilizing the historical data of projects. However, further work is required to fine-tune the weights set in the artifact traceability graph.Item Open Access Novel sampling strategies for experience replay mechanisms in off-policy deep reinforcement learning algorithms(2024-09) Mutlu, Furkan BurakExperience replay enables agents to effectively utilize their past experiences repeatedly to improve learning performance. Traditional strategies, such as vanilla experience replay, involve uniformly sampling from the replay buffer, which can lead to inefficiencies as they do not account for the varying importance of different transitions. More advanced methods, like Prioritized Experience Replay (PER), attempt to address this by adjusting the sampling probability of each transition according to its perceived importance. However, constantly recalculating these probabilities for every transition in the buffer after each iteration is computationally expensive and impractical for large-scale applications. Moreover, these methods do not necessarily enhance the performance of actor-critic-based reinforcement learning algorithms, as they typically rely on predefined metrics, such as Temporal Difference (TD) error, which do not directly represent the relevance of a transition to the agent’s policy. The importance of a transition can change dynamically throughout training, but existing approaches struggle to adapt to this due to computational constraints. Both vanilla sampling strategies and advanced methods like PER introduce biases toward certain transitions. Vanilla experience replay tends to favor older transitions, which may no longer be useful since they were often generated by a random policy during initialization. Meanwhile, PER is biased toward transitions with high TD errors, which primarily reflects errors in the critic network and may not correspond to improvements in the policy network, as there is no direct correlation between TD error and policy enhancement. Given these challenges, we propose a new sampling strategy designed to mitigate bias and ensure that every transition is used in updates an equal number of times. Our method, Corrected Uniform Experience Replay (CUER), leverages an efficient sum-tree structure to achieve fair sampling counts for all transitions. We evaluate CUER on various continuous control tasks and demonstrate that it outperforms both traditional and advanced replay mechanisms when applied to state-of-the-art off-policy deep reinforcement learning algorithms like TD3 and SAC. Empirical results indicate that CUER consistently improves sample efficiency without imposing a significant computational burden, leading to faster convergence and more stable learning performance.Item Embargo Aerobic oxidation of alcohols catalyzed by an efficient heterogeneous mixed metal hydroxide catalyst(2024-08) Bilgin, SuayAerobic alcohol oxidation is one of the important reactions of organic chemistry used both in academic research and in many industries such as medicine, cosmetics and food. Aldehydes, carboxylic acids and ketones resulting from alcohol oxidation are encountered as intermediates or final products. However, some oxidation methods developed to date have drawbacks such as expensiveness, being hazardous to human health and the environment, and having low sustainability. For this reason, Fe0.6Mn0.4(OH)x mixed metal hydroxide catalyst was developed in a collaborative work by the Ozensoy and Türkmen research groups by avoiding these drawbacks and was first used in the C-H activation reaction giving successful results. In this project, alcohol oxidation reactions were carried out using the same catalyst. Optimized reaction conditions were found by the model reaction of conversion from benzhydrol to benzophenone. Then, substrates with different electron withdrawing and donating groups attached to benzhydrol were tested along with benzyl alcohol and 1-phenylethanol derivatives. Finally, an attempt was made to obtain information about the working mechanism of the catalyst by performing a kinetic isotope effect experiment and various control experiments.Item Open Access Personality transfer in human animation: comparing handcrafted and data-driven approaches(2024-09) Ergüzen, Arçin ÜlküThe ability to perceive and alter personality traits in animation has significant implications for fields such as character animation and interactive media. Research and developments that use systematic tools or machine learning approaches show that personality can be perceived from different modalities such as audio, images, videos, and motions. Traditionally, handcrafted frameworks have been used to modulate motion and alter perceived personality traits. However, deep learning approaches also offer the potential for more nuanced and automated personality augmentation than handcrafted approaches. To address this evolving landscape, we compare the efficacy of handcrafted models with deep-learning models in altering perceived personality traits in animations. We examined various approaches for personality recognition, motion alteration, and motion generation. We developed two methods for modulating motions to alter OCEAN personality traits based on our findings. The first method is a handcrafted tool that modifies bone positions and rotations using Laban Movement Analysis (LMA) parameters. The second method involves a deep-learning model that separates motion content from personality traits. We could change the overall animation by altering the personality traits through this model. These models are evaluated through a three-part user study, revealing distinct strengths and limitations in both approaches.Item Open Access Adaptive control with LSTM augmentation: theory and human-in-the-loop validation(2024-08) İnanç, EmirhanThis thesis presents a novel adaptive control architecture that provides dramatically better transient response performance compared to conventional adaptive control methods. This is accomplished by the synergistic employment of a traditional Adaptive Neural Network (ANN) controller and a Long Short-Term Memory (LSTM) network. LSTM structures can take advantage of the dependencies in an input sequence, which helps predict uncertainty. We introduce a training approach through which the LSTM network learns to compensate for the deficiencies of the ANN controller in a closed-loop setting. This improves the system’s transient response and allows the controller to react to unexpected events quickly. This study also investigates the human-in-the-loop performance of the proposed control framework. Although the LSTM-augmented control method drastically improves the transient response, especially in the presence of significant and rapid uncertainty changes, its interactions with a human operator must be analyzed to ensure safe operation. First, a human pilot model is used to investigate the overall system’s behavior and explore the controller’s performance for a reference tracking task. Then, human-in-the-loop experiments are conducted to analyze how the system responds in the presence of an actual human operator in the loop. Through careful simulation studies, we demonstrate that this architecture improves the estimation accuracy on a diverse set of uncertainties. The overall system’s stability is analyzed via a rigorous Lyapunov analysis, and the proposed method is shown to be highly effective, as demonstrated through numerical simulations and human-in-the-loop experiments.