137091 research outputs found
Sort by
Phase Change Material Based Reconfigurable Phase Shifters and Filters for mmWave Antenna Arrays
No full textThis dissertation presents fundamental research of new radio frequency integrated circuit electronics, based on a phase change material (PCM), for antenna array technologies operating in the mmWave band (30–300 GHz). The results of this research include multiple PCM switches, phase shifters, and reconfigurable filters. These devices are shown to have advantageous characteristics such as wideband and ultrawideband performance, high power handling, low insertion loss, and rapid switching speed. Applications for these electronics include multifunction, multi-standard, and reconfigurable systems for 5G, 6G, Internet of Things (IoT), and Non-Terrestrial Networks (NTNs) at mmWave. The optimized integration of the phase change material, vanadium dioxide, into a five-step fabrication process is presented. The design, simulation, modeling, fabrication, and measurement of single pole single throw switches, single pole double throw switches, true time delay phase shifters, and a reconfigurable filter are presented with a discussion of the state of the art for comparable mmWave devices.Ph.D.Electrical and Computer Engineerin
Investigating the Oxidative Degradation of Aminopolymer Sorbents for Direct Air Capture (DAC)
No full textInvestigating the Oxidative Degradation of Aminopolymer Sorbent for Direct Air Capture (DAC)
Yoseph A. Guta
171 pages
Directed by Dr. Christopher W. Jones and Dr. Carsten Sievers
An investigation of the degradation mechanisms and ways of mitigation techniques was proposed. The stability of amine-functionalized sorbent (PEI/Al2O3) was investigated under conditions incorporating O2, CO2, and H2O at an intermediate temperature. Under the co-presence of CO2 (0.04%) and O2 (21%), substantial sorbent degradation is observed. The results show that adsorbed CO2 species, such as carbamic acid, catalyze the cleavage of the C-N bonds, leading to accelerated oxidative degradation. The CO2-induced degradation also remains significant under humid conditions, with its impact being higher in the absence of CO2. An in-depth experimental and computational analysis of the role of adsorbed CO2 species shows that chemisorbed CO2 species can enhance sorbent stability when present in high concentration (high CO2 loading). The high CO2 loading leads to crosslinking of amine chains, reducing the mobility of the aminopolymer chains and enhancing sorbent stability, while CO2 loadings below a certain threshold accelerate degradation due to the frequent interactions between adsorbed CO2 species and radical species. Mitigation against sorbent degradation was explored by incorporating radical scavengers and common hydrocarbon stabilizers with PEI/Al2O3 sorbents. The sorbent consisted of 4,4'-Bis(ɑ,ɑ-dimethylbenzyl)diphenylamine (BDDPA) showed enhanced oxidative stability under dry and humid CO2-free air and 0.04% CO2-air conditions at 70 and 120°C. The findings help obtain a fundamental understanding of the complex degradation mechanism of amine-functionalized sorbents in DAC processes and highlight the complexity each environmental component introduces in sorbent stability studies
Partitioning and Scheduling Framework with Dynamic Memory Estimation for Multi-Instance GPUs
No full textThe problem of partitioning and scheduling to effectively utilize Nvidia’s Multi Instance GPU (MIG) capabilities is quite challenging. On the one hand, tight partitions must be created to maximize concurrency and throughput; on the other, the memory needs of executing GPU processes must
be adequately met to avoid out-of-memory errors. This is exacerbated by the unique (dynamic) memory behavior of the modern ML workloads like LLMs, as well as by the peculiar constraints
on partition creation on the MIGs and choosing right configuration for maximizing concurrency.
This research proposes a comprehensive framework for solving the above challenges. It combines memory estimation analysis and scheduling to dynamically create and manage MIG parti-
tions, per the resource needs of GPU jobs. For general programs and ML workloads, we propose two scheduling schemes: one to minimize the number of repartitioning calls
at runtime, and a second that reconfigures the GPU partitions as per the need of the next GPU job in queue. This approach yields up to 6.20x throughput improvement and
5.93x energy improvement for general workloads; and we see 1.59x and 1.12x improvement to throughput and energy, respectively, for ML workloads on an A100 GPU.
Many workloads’ memory requirements, however, are quite challenging to analyze. State-of-the-art ML model estimation methods are ineffective for workloads that allocate dynamic
memory. To overcome this limitation, we design a time series-based profiling method that gathers memory allocation statistics during the initial part of the execution and then projects
future memory needs of the process. If the projected memory need is likely to exceed the allocated partition, the process is aborted and restarted on a larger partition. Early prediction
of memory needs is attempted to optimize delays in completing execution due to a restart. We leverage this technique on LLM workloads and show good improvements, including up to 1.43x throughput improvement and 1.11x energy savings. Lastly, we show that the framework is agnostic for MIG enabled GPUs, and can be adapted to newer generations of GPU micro-architectures without any changes
Reverse Osmosis Using Renewable Energy: System Design and Technoeconomic Optimization for Distributed Desalination
No full textFreshwater use in energy generation, manufacturing, and agriculture has resulted in a significant increase in global water demand. Desalination is a technology solution that augments freshwater reserves by treating saline sources using an energy input. Reverse osmosis (RO) is a mature membrane-based technology for seawater desalination, but centralized systems consume significant amounts of energy that contribute to the cost and carbon footprint of the product water. This has led to an emerging interest in coupling RO with renewable energy sources (RES) at a distributed scale to treat inland saline sources, like brackish groundwater or oil and gas produced water. However, the intermittency of solar photovoltaics (PV) and wind is a challenge given that conventional RO is a constant pressure process designed to operate continuously and maximize uptime. Batch Reverse Osmosis (BRO) is a novel variable pressure process that has been demonstrated to operate at higher water fluxes, higher water recovery, and with lower energy input. These attributes make it suitable for a renewably paired system, but this has yet to be investigated. This research (i) develops an analysis framework for RO driven by intermittent RES, and (ii) evaluates the performance of BRO compared to conventional RO.
The system design comprises four subsystems: energy generation (RES), energy storage (battery), RO (conventional and BRO), and water storage (tank). The RO energy consumption and flux are obtained using the solution-diffusion model. System operation over one year is simulated, and techno-economic modeling and optimization techniques are applied to identify subsystem sizes that yield the lowest levelized cost of water (LCOW). This framework is also applied to other energy resources (wind, hybrid, and grid-connected), feed salinities (brackish and produced water), and water demand profiles (flat and agricultural). For PV-RO with 33.3 ppt feed, the analysis reveals that the higher flux operation of BRO can (i) increase direct use of renewables 40%, (ii) reduce energy storage needs 16×, and (iii) reduce curtailed electricity 6%. BRO can also (iv) reduce brine production 1.19× by operating at higher recovery, or (v) reduce membrane area and energy system size 1.10× by operating at a lower specific energy consumption. These attributes result in a total LCOW reduction of 1.01−1.08× for BRO compared to conventional RO. At brackish water salinities (3 – 10 ppt), BRO is less energy efficient than conventional RO but can reduce brine production ~2×. The case studies also reveal that wind-RO and hybrid-RO have 1.03× and 1.14× lower costs than PV-RO, respectively. Overall, this work identifies key design tradeoffs in RES-driven RO systems to enable sustainable water production.M.S.Mechanical Engineerin
Regenerative Engineering and Rehabilitation Strategies for Bone Regeneration and Functional Restoration
No full textCritical bone defect injuries present a significant clinical burden. The current standard of care often results in nonunion and requires multiple revisions surgeries. Even if rescue attempts are eventually successful, the long healing period ultimately leaves patients with long-term disability and chronic pain. Bone healing relies on a precise interplay of biochemical and mechanical cues to guide the healing process. However, the current biological treatments have many limitations and negative side effects, and current bone stabilization methods prevent any mechanical stimulation within the defect region. The overarching hypothesis of this thesis is that an osteogenic therapeutic will enhance bone formation and bridging while the early induction of mechanical stimulation via rehabilitation will enhance bone formation and promote the functional recovery of the injured limb. Therefore, the overall objective of this thesis is to establish a multi-faceted approach to treat critical bone loss injuries. This approach integrates an osteogenic therapeutic to stimulate bone regeneration with a rehabilitation regimen beginning early in the healing process to target functional restoration. This will be achieved through four specific aims: 1) engineering an osteogenic small molecule therapeutic targeting the BMP pathway to induce bone regeneration, 2) establishing the impacts of early induction of mechanical loading on bone healing, 3) developing a mouse bone defect model to determine the impacts of early induction of mechanical loading on functional recovery, and 4) establishing a combined therapeutic strategy consisting of an osteogenic therapeutic to induce bone formation and early rehabilitation to enhance functional recovery. Overall, this thesis uncovers novel osteogenic therapeutics, and furthers our understanding on the effects of mechanical stimulation through rehabilitation regimens on bone healing and functional restoration. The knowledge attained by this thesis provides the groundwork for the development of integrative therapeutic strategies that target functionality alongside tissue regeneration.Ph.D.Biomedical Engineerin
Propagation Dynamics in Cardiac Tissue: From Alternans Formation to Defibrillation
No full textCardiovascular disease is the leading cause of death worldwide, and many fatal events are caused by cardiac arrhythmias. Clinically, T-wave alternans is an important marker of arrhythmia risk. Electrical cardioversion is the standard shock-based treatment used to restore normal rhythm clinically, but current devices rely on high-energy shocks, which can be painful and may cause tissue damage. To develop safer, low-energy alternatives, we need to better understand alternans dynamics and how external electric fields interact with unstable electrical activity in the heart.
This thesis focuses on two related questions. First, we study how spatial coupling in cardiac tissue affects alternans and wave stability by introducing fractional diffusion, a modeling approach that captures the physiological characteristics of structural heterogeneity. Using the Beeler–Reuter model and the Fenton–Karma model, we identify the conditions under which fractional diffusion only changes the spatial length scale of alternans without altering alternans dynamics. This helps tackle the alternans size mismatch problem between experiments and numerical simulations, improving the predictive power of simulations.
Second, we simulate fibrillation in two-dimensional tissue and study the efficacy of rotational electric fields for defibrillation. Through voltage maps and phase singularity analysis, we present examples demonstrating that rotational fields can defibrillate better than static electric field fields when applied at the same amplitude, suggesting that the use of rotational electric field may be a more effective low-energy approach
Development of a UAV-Mounted System for Physical Interception of Aerial Vehicles
No full textGiven the proliferation of low-cost, highly configurable unmanned aerial vehicles (UAVs), counter-UAV systems have arisen as an essential means for maintaining aerial security. However, current solutions present a trade-off: kinetic systems that physically capture vehicles require complex post-capture controllers, while non-kinetic methods like jamming pose collateral risk to the environment without capture. This research investigates the feasibility of a novel, UAV-mounted retractable cage mechanism designed to offer a stable, non-destructive, physically secure capture method. The objective is to demonstrate the mechanism’s capability at autonomously intercepting fixed aerial targets. This research presents the novel mechanism as well as a dynamic Simulink model to estimate the vehicle’s behavior and flight controller performance. Through experimental testing, the feasibility of the capture mechanism and autonomous algorithms have been validated. However, the intervehicle coordination between the capture mechanism and moving targets still remains a challenge. Future work should test the feasibility of the retractable cage mechanism when capturing moving targets
A Surrogate-Assisted Online Adaptive Reinforcement Learning and Approximate Bayesian Computation (OARL-ABC) Method for Calibration of Digital Twins
No full textDigital twin technology has become a cornerstone in modern industry and research, providing virtual replicas of physical systems that enable real-time condition monitoring, future state simulation, and design optimization capabilities over a product's entire life-cycle. The accuracy and reliability of digital twins depends critically on the calibration process, which aligns the digital model with real-world data. As physical systems evolve over time, digital twins must be continuously recalibrated to remain accurate representations of their connected physical counterparts. Recent trends in digital twin applications have demanded more complex model forms and more stringent recalibration time-frames, creating an urgent need for improved calibration methods that can handle this increased complexity while maintaining computational efficiency.
This thesis addresses these challenges by developing and validating a new surrogate model assisted online adaptive reinforcement learning and approximate Bayesian computation (OARL-ABC) method for the calibration and validation of digital twins. The approach builds upon a reinforcement learning-based calibration framework that performs model selection via reinforcement learning techniques and parameter calibration via Bayesian inference methods, specifically approximate Bayesian computation (ABC). The method employs a hybrid Bayesian reinforcement learning calibration framework that combines the adaptability and efficiency of reinforcement learning for optimizing complex, dynamic systems with the uncertainty quantification and updating capability of Bayesian inference methods. This combination integrates the principled uncertainty handling of Bayesian inference with the adaptive learning capabilities of reinforcement learning, making it particularly well-suited for the calibration of highly complex digital twins.
To enhance the efficiency of Bayesian reinforcement learning, this thesis integrated surrogate modeling to provide computationally efficient approximations of complex models for more rapid evaluation in the learning process. This surrogate-assisted OARL-ABC method successfully reduced the computational intensity of the sampling requirements by utilizing multiple surrogate model representations of real systems. Through investigation of various surrogate modeling techniques, Bayesian network models were identified and implemented as the optimal choice in this context, as these models maintain the Bayesian framework's ability to manage uncertainty while significantly reducing computational demands, thereby accelerating the calibration process without compromising accuracy.
The resulting OARL-ABC method demonstrated increased adaptability to different model forms, improved efficiency in complex applications, and robust capability to quantify inherent problem uncertainties. The method's ability to capture complex system behaviors, account for uncertainties, and perform continuous learning in a single integrated loop achieved substantial reductions in computational cost compared to conventional approaches for higher-order models.
The effectiveness of this calibration method was validated through application to a rotordynamic system with both physical and digital counterparts that could be scaled in complexity and manipulated to represent different stages of a product's life-cycle. A machinery fault simulator (MFS) test rig provided the physical platform, designed as a flexible simulator for various types of rotating machinery faults that could be easily reconfigured to represent different systems of increasing complexity and degradation throughout the product life-cycle. The MFS uses an electric motor attached to variable disks, weights, bearings, and a shaft to simulate different fault conditions found in real-world rotating machinery such as turbines or compressors. By systematically varying the distribution of weights around the axle, multiple imbalance fault conditions were generated and matched to their respective total output vibration signatures, measured by accelerometers mounted to the bearing housings. These physical experiments were paired with a high-fidelity finite element rotor dynamics models created in Python using Rotordynamic Open-Source Software (ROSS), an open-source rotordynamics library serving as the digital representation of the simulated system to capture the complex dynamics of rotating systems.
The validation process constructed surrogate model representations of various experimental configurations, from which the reinforcement learning algorithm selected the best representation for each state of the physical system while Bayesian inference calibrated the model parameters. The performance of the proposed method was then systematically compared in this rotordynamic context against benchmark performances using conventional ABC calibration, conventional OARL-ABC without surrogate model assistance, and OARL-ABC using Bayesian network surrogate models across increasingly complex rotating machine simulations.
Through this comprehensive rotordynamic testing campaign, this research successfully demonstrated the improved performance and scalability of the surrogate assisted OARL-ABC calibration method for real-world digital twin systems. The approach delivered substantial improvements in calibration efficiency while enhancing robustness and reliability when applied to increasingly complex digital twin models with demanding recalibration requirements. These results establish a foundation for more effective and widespread application of digital twin technology in modern industry and research, enabling continuous model adaptation and uncertainty quantification at computational costs that make real-time calibration practically feasible for complex systems
Assortment Optimization under Customer-driven Substitution
No full textRetailers face growing complexity in assortment planning due to the rise of omnichannel retailing, the proliferation of product lines, and increasing uncertainty in supply chain operations. Motivated by these challenges, this dissertation focuses on two distinct problem contexts. The first is the transformation of offline stores into experiential showrooms, which requires new models for designing in-store assortments that shape customer purchase decisions beyond immediate product availability. The second is the need to integrate assortment decisions with upstream supply availability and production capacity constraints, as well as downstream fulfillment dynamics, to maintain responsiveness under volatile and uncertain operating conditions.
Customer substitution behavior—the propensity of customers to select alternative products when preferred items are unavailable—serves as a critical lever in both contexts, enabling retailers to better align supply with demand. However, substitution behavior is often simplified or underrepresented in supply chain optimization models. This dissertation addresses this gap by introducing a new form of substitution behavior in a novel retail context and developing frameworks that explicitly incorporate customer-driven substitution effects into assortment planning through optimization- and simulation-based decision-support models. These models are applied to proprietary industry data and validated through realistic case studies.
The first module of the dissertation (Chapter 2) introduces the showroom assortment optimization problem, a novel approach for showcasing optimization in omnichannel retail networks in which offline stores operate as experiential showrooms rather than traditional fulfillment centers. In this setting, customers interact with products in-store before making online purchase decisions, and substitution behavior is shaped by in-store experiences rather than immediate product availability. We define the showcasing optimization problem as selecting the assortment of products that maximizes the purchase confidence of an average customer following a store visit while adhering to capacity constraints. We formulate the underlying optimization problem as a mixed-integer program (MIP) that captures how customers gain purchase confidence through surrogate products—items that represent or stand in for other products—thereby modeling the customer substitution effect in the showroom environment. Using data from an industry partner, we demonstrate the practical applicability of our model in quantitatively designing effective showcasing strategies to improve customer purchase confidence and stimulate sales.
The second module (Chapters 3–4) addresses assortment planning under supply and demand uncertainty for a make-to-stock manufacturer–retailer operating in capacity-constrained production and fulfillment settings. Chapter 3 examines the profit-maximizing multi-period assortment planning problem. We develop a stochastic choice-based optimization model that endogenizes customer substitution behavior through a rank-based choice model with small consideration sets, enabling computationally tractable approximations of stockout-driven substitution probabilities. This model extends traditional assortment planning by explicitly linking assortment decisions with upstream supply availability and production capacity constraints, as well as downstream fulfillment performance. To solve realistically sized instances, we employ a rolling-horizon framework with multi-period lookahead and a two-stage stochastic program, where a Benders decomposition separates master assortment and sourcing decisions from scenario-specific subproblems that allocate production, inventory, and fulfillment over time. Case study results with a North America–based industry partner demonstrate substantial improvements in profit and service levels compared to baseline static or myopic tactical plans, with benefits amplified under volatile supply and demand conditions.
Chapter 4 develops a hybrid simulation–optimization framework to refine tactical assortments through operational adjustments at finer review intervals. Tactical plans, set during the sales and operations planning process, fix supplier and material commitments but allow limited assortment changes within operational review periods. The framework integrates a mixed-integer linear program (MILP) with a high-fidelity multi-agent system (MAS) that simulates detailed operational dynamics, including stochastic demand, inventory flows across multiple echelons, and demand fulfillment processes. In each review period, the MILP proposes state-contingent adjustments, which the MAS evaluates under realistic operating conditions. A feedback-driven neighborhood search iteratively improves solutions to ensure both operational feasibility and profitability. This matheuristic approach enables alignment between tactical assortment planning and operational execution, capturing stochastic and behavioral effects not fully represented in optimization models. Case study experiments show measurable gains in profit and service levels over static tactical plans, underscoring the importance of operationally informed assortment adjustments.
Collectively, this work advances the integration of customer choice modeling with assortment planning across diverse retail contexts. In the showroom setting, the proposed models capture how surrogate products influence in-person purchase confidence, enabling retailers to design in-store assortments that shape customer decisions beyond immediate product availability. In operational supply chain environments, the frameworks integrate substitution-driven demand modeling with capacity, procurement, and fulfillment constraints to hedge against uncertainty, improve demand fulfillment, and enhance resilience. These contributions demonstrate how substitution can serve as a unifying lever to align customer behavior with supply and operational realities. The models offer actionable guidance for retailers seeking to design assortments that balance profitability, customer experience, and operational robustness in increasingly complex and volatile environments