Virginia Tech - Wake Forest University School of Biomedical Engineering & Sciences
VTech Works (Virginia Tech)Not a member yet
122049 research outputs found
Sort by
Advanced Science
No full textBy integrating the principles of kirigami cutting and data-driven modeling, this study aims to develop a personalized, rapid, and low-cost design and fabrication pipeline for creating body-conformable surfaces around the knee joint. The process begins with 3D scanning of the anterior knee surface of human subjects, followed by extracting the corresponding skin deformation between two joint angles in terms of longitudinal strain and Poisson's ratio. In parallel, a machine learning model is constructed using extensive simulation data from experimentally calibrated finite element analysis. This model employs Gaussian Process (GP) regression to relate kirigami cut lengths to the resulting longitudinal strain and Poisson's ratio. With an R2 score of 0.996, GP regression outperforms other models in predicting kirigami's large deformations. Finally, an inverse design approach based on the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) is used to generate kirigami patch designs that replicate the in-plane skin deformation observed from the knee scans. This pipeline was applied to three human subjects, and the resulting kirigami knee patches were fabricated using rapid laser cutting, requiring less than a business day from knee scanning to kirigami patch delivery. The low-cost, personalized kirigami patches successfully conformed to over 75% of the skin area across all subjects. The kirigami-inspired, machine-learning-driven design and fabrication pipeline presents a balanced trade-off between conformability performance and cost for personalizing wearables, thus establishing a foundation for a wide range of new functional devices.Published versio
One Continuous Side: Rethinking Followership with the Möbius Strip
No full textWhat if leading and following aren’t opposites, but the same continuous skill? Using the Möbius Strip as a hands-on metaphor, this session reimagines leadership as shared, dynamic, and courageous—no matter where you stand
Constructing Commonsense Knowledge Graph for Persona Consistency
No full textEnsuring consistent persona in interactive AI systems presents a significant challenge, especially in diverse application scenarios ranging from virtual assistants to customer service bots. Such capability is often constrained by the system's understanding of direct and explicit persona conflicts. Traditional approaches primarily focus on detecting discrepancies between machine responses and its predefined profile, or the contextual inconsistencies between the responses at the semantic level rather than the persona level. Due to the lack of a comprehensive persona-specific Commonsense Knowledge Graph, some indirect and implicit persona inconsistencies between machine responses can hardly be identified. In this paper, we build the first persona commonsense knowledge graph (PersonaKG), based on which we then construct a large-scale persona consistency dialogue dataset (PersonaCOM) containing both explicit and implicit persona conflicts between machine responses. With the guidance of the persona commonsense knowledge, we propose a Recognize-Rewrite framework (R2) which first recognizes the responses that are inconsistent in persona with the previous responses, and then rewrites them into consistent ones. The empirical study demonstrates that utilizing R2 method on PersonaCOM with PersonaKG results in a significant improvement of 12.20% in automatic metrics and 10.09% in manual evaluation compared to not using the R2 method and PersonaKG.Published versio
Subband Stitching for D-Band Measurements
No full textMaster of ScienceNew technologies are increasingly demanding electrical communication systems with higher data rates and lower latency. To satisfy these needs, researchers are exploring the feasibility of lesser used high-frequency bands for the deployment of future systems. As part of this research, channel measurement campaigns are being conducted to characterize the channel over wide ranges of frequencies. In this thesis, we explore how a technique known as subband stitching can be combined with an existing channel measurement procedure in order to increase the range of frequencies that a system can measure. We provide an analysis of this subband stitching technique and present detailed simulation results evaluating it's accuracy. Additionally, we implement this design on a real-world system in order to obtain wide bandwidth channel measurements
Rider Insights on Motorcycle Safety Tech: What Drives—or Blocks—Adoption of ABS, MSC, and AEB
No full textAdvanced Rider Assistance Systems (ARAS) have multiple benefits to motorcycle riders, auto manufacturers, individual states, and any person interested in reducing roadway fatalities. ARAS technologies are a promising approach to reducing fatalities by preventing the crashes that cause them. Three technologies were selected for study utilizing a survey-based approach to understand riders’ attitudes toward their use. Anti-lock brakes (ABS) represent an existing and mature technology likely encountered by many riders. Motorcycle Stability Control (MSC) is a technology present on several high-end motorcycles and uses braking to stabilize the motorcycle in a curve. Automatic Emergency Braking (AEB) is a nascent technology in the motorcycle world but has existed since the early 2000s in the passenger vehicle domain. This study surveyed 1,391 licensed riders and sought to accomplish four objectives related to understanding attitudes toward ARAS, gaining specificity behind these attitudes, understanding barriers, and overcoming those barriers.
Overall, motorcycle riders recognize that ARAS improve safety, but adoption is slowed by cost, misunderstanding, and concerns about riding identity. These barriers are solvable. Through hands-on experiences, education, and incentives, stakeholders can accelerate ARAS adoption, improve rider safety, and drive market growth.National Surface Transportation Safety Center for Excellenc
Optics Letters
No full textWe propose encrypted fringe projection profilometry (E-FPP), a co-keyed phasecode framework that embeds encryption directly into the measurement process of fringe projection profilometry (FPP). Two random phase fields and orthogonal temporal codes jointly encode the projected sequence, binding phase retrieval to legitimate key pairs. The method preserves standard three-step FPP operation while enabling key-dependent 3D reconstruction. Experiments verify accurate authorized reconstruction and strong resistance to single- and multi-frame inference, providing a privacy-preserving solution for optical metrology.Submitted versio
A stable isotope and macro-charcoal sediment record spanning the Pleistocene-Holocene transition from Maple Pond in the southern Shenandoah Valley, Virginia, USA
No full textThis study presents a paleoenvironmental analysis of a 147-cm sediment core from Maple Pond, a sinkhole pond within the Shenandoah Valley Sinkhole Pond complex at the base of the Blue Ridge in central Virginia. The results document more than 15,000 years of continuous sediment accumulation, fire history, and environmental variability from the Late Pleistocene to the present. Stable carbon isotope ratios (δ13C: -29‰ to -27‰) and atomic C:N ratios (11-36) indicate terrestrial C₃ vegetation as the primary source of organic matter during the Late Pleistocene and early Holocene (15,000–8,000 cal yr BP), followed by a transition toward nitrogen-rich aquatic and wetland inputs during the mid-to-late Holocene. Loss-on-ignition analysis records organic content rising from ~10% in basal sediments to 35–40% in the upper core, reflecting a transition from minerogenic deposition to sustained organic accumulation. Macroscopic charcoal concentrations reveal a mid-Holocene maximum (40–49 cm depth; ~6,000–5000 cal yr BP), attributed to increased regional fire activity, possibly related to warmer drier conditions during the Mid-Holocene Maximum and a shift to fire-adapted forest (oak, hickory, pine). A smaller charcoal peak between 120 and 130 cm may relate to rapid warming at the end of the Younger Dryas, a shift back to cold conditions around 12,000 years ago. The predominance of lower charcoal abundance (<100 fragments/cm3) characterizing the majority of the sequence suggests a long record of frequent, low-intensity fire regimes characteristic of presettlement fire-adapted Appalachian forests. Geochemical and fire history patterns at Maple Pond demonstrate broad consistency with paleoenvironmental reconstructions from nearby sites (Browns Pond, Spring (aka Hack) Pond, Twin Pond and Cranberry Glades), supporting the interpretation that sinkhole ponds and wetlands throughout the Shenandoah Valley underwent similar environmental evolutions across major Late Quaternary climatic transitions.Master of ScienceUnderstanding how past landscapes and ecosystem changes can inform predictions of future climate responses requires the examination of long-term environmental records. This study analyzes a 147 cm sediment core from Maple Pond, a sinkhole wetland in Augusta County, Virginia, within the Shenandoah Valley. Chemical analysis and charcoal measurements from sediments spanning more than 15,000 years document environmental history from the end of the last Ice Age to the present. Results reveal changing plant communities through time. Between 15,000 and 8,000 years ago, forests dominated the surrounding landscape, and most organic material in the pond emerged from land plants. Over the past 8,000 years, wetland and aquatic plants have contributed increasingly to sediment composition, indicating a transition to a more productive wetland system. Charcoal preserved in sediments was also examined to understand the history of fires in this area. Consistent macrocharcoal levels throughout most of the sequence likely reflect low-intensity fires, likely occurring at intervals of several hundred to a thousand years. A notable charcoal peak between 5,000 and 6,000 years ago may indicate a period of increased fire activity associated with warmer, drier conditions during the mid-Holocene. Patterns observed at Maple Pond parallel those from nearby sites, suggesting that sinkhole wetlands across the Shenandoah Valley experienced comparable environmental changes over millennia. This research highlights the importance of small wetlands serving as natural records of past environmental conditions
SLED: A Speculative LLM Decoding Framework for Efficient Edge Serving
No full textThe growing gap between the increasing complexity of large language models (LLMs) and the limited computational budgets of edge devices poses a key challenge for efficient on-device inference, despite gradual improvements in hardware capabilities. Existing strategies, such as aggressive quantization, pruning, or remote inference, trade accuracy for efficiency or lead to substantial cost burdens. This position paper introduces a new framework that leverages speculative decoding, previously viewed primarily as a decoding acceleration technique for autoregressive generation of LLMs, as a promising approach specifically adapted for edge computing by orchestrating computation across heterogeneous devices. We propose SLED, a framework that allows lightweight edge devices to draft multiple candidate tokens locally using diverse draft models, while a single, shared edge server verifies the tokens utilizing a more precise target model. To further increase the efficiency of verification, the edge server batches the diverse verification requests from devices. This approach supports heterogeneous devices and reduces server-side memory footprint by sharing a single upstream target model across devices. Our initial experiments with Jetson Orin Nano, Raspberry Pi 4B/5, and an edge server equipped with 4 Nvidia A100 GPUs indicate substantial benefits: ×2.2 higher system throughput, ×2.8 higher system capacity, and better cost efficiency, all without sacrificing model accuracy.Published versio
Aerodynamic Enhancement and Reduced Order Modeling of Vertical Axis Wind Turbines
No full textVertical-axis wind turbines (VAWTs) are recognized as a viable solution for wind energy harvesting. This study discusses VAWT performance in different aspects. First, using computational fluid dynamics (CFD) simulations, the impact of various deflector angles as an auxiliary augmentation on turbine efficiency is examined. Specifically, the deflector orientation angle effect on the dual-rotor straight blade vertical-axis wind turbine (DR-SBVAWT) performance is conducted. Two-dimensional transient simulations were performed for this parametric study. The results demonstrate that deflector implementation boosts the DRSBVAWT self starting capabilities and enhances overall performance. With a vertical deflector (β = 0◦from the y-axis) yields the best performance, providing the highest efficiency and power output. In contrast, a horizontal deflector (angle of β = 90◦ counterclockwise from the y-axis) shows minimal impact on the turbine's performance, suggesting that further angle variations do not significantly enhance the system. Moreover, sensitivity analysis was performed to evaluate the impact of small changes in the deflector orientation that shows β = 0◦ holds the best orientation and small angle variations in deflector orientation show minimal impact on the overall performance of the turbine. In steady-state conditions, the vertical deflector angle increases the tip speed ratio (TSR) of the DR-SBVAWT performance by 11.5% compared to a conventional DR-SBVAWT without the deflector. Additionally, this vertical configuration achieved a 30.15% increase in efficiency at TSR = 2.5, showing its effectiveness in improving overall aerodynamic performance. This parametric study overall provides valuable insights into the optimal deflector angle configuration as an auxiliary augmentation system for dual-rotor vertical axis wind turbines, contributing to the design optimization and improved performance of wind energy systems.
Secondly, utilizing the same method of a 2D transient unsteady Reynolds-averaged Navier– Stokes (URANS) numerical simulations, different clustering scenarios of DR-SBVAWT for farm design are studied. Twelve clustering configurations include vertically aligned pairs and staggered clusters of three turbines at different inter-turbine distances, to evaluate their impact on power capture and land usage for the DR-SBVAWT, are investigated. Performance indices, namely, total power coefficient and improvement relative to standalone turbines, are analyzed. Wake effects are qualitatively discussed through detailed velocity contour plots of the wind field. Results show that a DR-SBVAWT turbine arrangement can enhance wind farm performance by approximately 25% for two turbines and about 20% for three staggered turbines, with required spacing of 1.5D and 2.5-3D, respectively. The study overall provides critical insights into the optimal placement and configuration of DR-SBVAWTs for maximizing energy output while minimizing land usage, offering guidance for the design of more efficient VAWT farms. Furthermore, it offers guidance on mitigating destructive interference from upstream turbines, enabling the optimization of multi-turbine layouts so that each unit operates under stable flow conditions.
Finally, a reduced order model (ROM) is investigated for VAWT. A novel and robust CFDROM framework is built to evaluate the complex flow field behaviour of a single rotor 3-blade VAWT. A data-driven framework was developed following high-fidelity transient simulations conducted using the URANS simulations within a finite volume method (FVM) framework in ANSYS Fluent. A snapshot-based proper orthogonal decomposition (POD) reduced-order model was developed. Additionally, a domain decomposition strategy was implemented to analyze the flow behavior across the interface between subdomains (the rotor domain and its surrounding domain) of VAWT. To ensure accurate enforcement of interface continuity conditions, the coupling between the inner and outer domains was achieved through a tunable proportional controller (computational gain). Results reveal that domain decomposition approach coupled with ROMs enhances the robustness of the computational cost while maintaining acceptable accuracy. By adding the computational controller, it helps to improve the coupling of the two ROMs, ensuring more efficient integration with fewer mismatches at the interface between the decomposed domains. The data-driven novel framework provides a critical stepping point to extending research to include more complexity of flow dynamics of VAWTs. This proposed framework proves a speedup on the order of one for 300 snapshots, or on about order of four for complete transient simulations (20 seconds) of fixed VAWT.
Overall, given the greater viability and applicability of VAWTs in different setups compared to other turbine types, these three detailed studies of dual-rotor design, farm design, and ROM for VAWT, offer a comprehensive framework for understanding VAWT aerodynamics and provide more efficient, high-fidelity design and analysis that benefits in varied wind environments.Doctor of PhilosophyWind energy is a clean and renewable way to generate electricity, but traditional wind turbines can be expensive, noisy, or difficult to install in different setups, such as urban environments. Unlike traditional horizontal axis wind turbines (HAWTs), vertical axis wind turbines (VAWTs) offer a flexible alternative because they can capture wind from any direction, operate closer to the ground, and have simpler designs. However, many VAWTs struggle to start spinning in low winds and may produce less power than expected, compared to HAWTs. This dissertation explores new ways to improve VAWT performance, focusing on recent designs of dual-rotor VAWT aiming to enhance the self-starting ability and boost overall efficiency. Using advanced computational tools, the study investigates how a deflector design, used as an auxiliary augmentation device that acts as an enhanced wind collector, boosts the self-starting capability of dual-rotor straight-blade vertical axis wind turbines (DR-SBVAWT). Results show that vertical deflector plates, as a simple design, can improve starting torque and overall power output, while keeping the whole system simple in design and manufacturing. Furthermore, the DR-SBVAWT shows advantages for the wind farm setup with less land usage. Dual-rotor design requires less space for wake recovery, which impacts the neighboring turbines within the wind farm. The study also provides a substantial guide for the best farm layout when using DR-SBVAWT. These studies offer practical insights for the wind industry, facilitating efficient deployment in multiple settings, including urban, onshore, and offshore environments. Finally, given a significant computational cost of VAWT, a novel data-driven robust framework is developed for VAWT modeling with accurate and less computational costs. It couples a reduced-order model and domain decomposition that provides a robust model reduction for turbulent fluid flow and for VAWT, particularly. This framework is crucial for routine designs and optimization for high computational costs, such as VAWT. Overall, findings provide valuable insights and give a practical path toward more efficient and reliable wind turbine solutions, making clean energy more applicable in different setups, accessible to communities, and helping reduce dependence on fossil fuels
Aircraft System Identification Approach for Control Surface Fault Diagnosis
No full textModern fault detection and diagnosis (FDD) methods are critical to maintaining flight vehicle safety. This thesis presents a model-based FDD approach for identifying control-surface loss of effectiveness on a small, fixed-wing research aircraft. The work considers several real-time system identification methods to estimate changes in control effectiveness and provide fault information to a fault-tolerant control allocation framework. A baseline aero-propulsive model for an experimental aircraft was developed from flight-test data to establish nominal control-effectiveness parameters used to compare fault diagnosis methods. Five real-time estimation methods were formulated and evaluated: exponentially weighted recursive least squares in the time- and frequency-domain, two Lyapunov-based adaptive parameter estimation methods with exponential or finite-time convergence guarantees under a persistence of excitation condition, and an augmented-state extended Kalman filter. These methods were applied to flight data containing an artificially injected stuck left-aileron fault, implemented through a custom maneuver injection capability, with multisine excitation inputs applied to the control effectors. The estimated control-effectiveness parameters associated with the faulted surface displayed an immediate response to the failure and a clear trend towards zero, while the parameters corresponding to healthy effectors remained near nominal values. The resulting estimates were used to construct a time-varying health matrix that scales the nominal control-effectiveness matrix, producing a fault-weighted representation suitable for control allocation and supporting the objective of fault hiding. Overall, this work advances in-flight fault diagnosis by providing real-time parameter estimation for fault-tolerant control allocation, enabling redistribution of control authority to support flight operations.Master of ScienceIt is essential for aircraft to reliably recognize and respond when a control surface is no longer performing as intended. If an aileron, elevator, or rudder becomes stuck or loses effectiveness, the aircraft can still fly safely, provided the control system can quickly identify what changed and redistribute control effort to the remaining healthy surfaces. This thesis demonstrates how real-time modeling can be used to estimate the in-flight health of each control surface on a fixed-wing research aircraft using flight-test data. Initially, a baseline model of the aircraft was developed using test maneuvers that excite the control effectors and characterize the aircraft motion in nominal conditions. Then, multiple real-time parameter-estimation methods were evaluated using flight data containing an intentionally injected fault in the left aileron implemented through custom flight software. The methods detected a loss of the left aileron's effectiveness while indicating that the other surfaces remained healthy. These estimates were then used to construct a control-surface health matrix that could be implemented into a fault-tolerant control allocation algorithm, enabling the aircraft to shift control authority away from the degraded surface and maintain its nominal control performance