Privacy Done Differentially
The Forest Inventory and Analysis (FIA) program of the USDA reaps several benefits from publishing its plot data. For instance, disclosure allows third-party researchers to help further its mission of monitoring forest trends in the U.S. However, privacy obligations complicate data sharing. To protect the location of its plots, the FIA must first randomly jitter plot coordinates before mapping plots to important auxiliary information. This procedure alters the statistical patterns of the data, which has implications for small area estimation. The goal of this project was to help the FIA determine whether a novel technique, Differential Privacy (DP), enhances both data privacy and accuracy. Of particular interest to the team was whether we could achieve these benefits by adding random noise to each plot coordinate and its corresponding auxiliary data. Our findings indicate that while we can theoretically design a DP mechanism to compute microdata, such a method is taxing in terms of utility. Therefore, we recommend that the FIA conduct research into DP computations that generate synthetic microdata that preserve trends in the original data.

Torsion, Tensile, and Impact Performance of Additively Manufactured PA6 Carbon Fiber and Glass Fiber Composites

Additive manufacturing (AM) using fused deposition modeling (FDM) is increasingly applied to fiber-reinforced polymers, yet comprehensive experimental evaluation of their mechanical behavior remains limited. This study expands prior work on carbon-fiber (CF) and glass-fiber (GF) reinforced nylon by investigating the influence of print parameters, annealing conditions, and filament type on the tensile, fatigue, and torsional performance of reinforced and pure nylon specimens. Test coupons were iteratively fabricated using a dual-nozzle FDM system, with process refinements implemented to address support-material adhesion, filament moisture, and print variability. Mechanical properties, including tensile strength, Young’s modulus, ductility, and torsional resistance, were extracted through standardized testing and custom data-analysis methods. Results indicate that appropriate annealing improves consistency in CF and GF composites, though GF specimens exhibited higher print failure rates. Pure nylon demonstrated high ductility and irregular fracture behavior, likely influenced by moisture absorption and thermal history. Measured elastic moduli were generally lower than manufacturer specifications, suggesting sensitivity to testing conditions and strain measurement methods. Torsional testing showed the greatest reproducibility in pure nylon, while chopped carbon-fiber reinforcement provided the strongest resistance to torsional deformation and fracture. Overall, significant variability across all materials highlights the strong dependence of mechanical performance on processing conditions. These findings contribute to improved understanding and optimization of FDM-printed nylon composites for structural applications.

Comparative Validation of Gel-Based and Gel-Less Electroencephalography Systems for Future Neurophysiological Studies

This study compares gel-based and gel-free electroencephalography (EEG) systems
using the OpenBCI Cyton + Daisy board, a low-cost, open-source platform increasingly utilized in neurotechnology research. While gel electrodes are known for superior signal quality, they require extensive setup and can cause user discomfort. Gel-free systems offer greater convenience, but may introduce higher noise and signal instability. Prior studies have assessed these trade-offs using clinical-grade hardware, but few have evaluated performance specifically within OpenBCI systems.

Five healthy participants completed EEG tasks including resting-state (eyes-open/closed) and motor tasks (finger tapping, spiral drawing) using both caps. Signal quality was evaluated through alpha-to-beta (A:B) ratios, signal-to-noise ratio (SNR), impedance levels, and independent component analysis (ICA) -derived artifact proportions. Usability was assessed through setup time, impedance stability, and a user comfort survey.

The gel-based system showed stronger and more consistent A:B elevation during eyes-closed conditions, though group differences were not statistically significant. SNR values were similar between systems (gel: 13.63 ± 22.46 dB; gel-free: 19.22 ± 17.55 dB), with one gel outlier likely due to setup error. Artifact proportions were higher in gel data, while impedance was lower and more stable compared to gel-free. Survey results indicated similar comfort, with minimal post-use discomfort or cleanup issues.

Despite a small sample size, both systems effectively captured relevant neural signals.
Gel-based systems showed marginally greater consistency, while gel-free systems offered setup advantages. These findings provide early insights for selecting EEG systems based on specific research needs. Future work will explore additional metrics such as spectral stability and topographical reliability across broader populations.

Numerical Analysis of Vane Shear Test
The Vane Shear Test (VST) is widely used in geotechnical engineering to determine the undrained shear strength of soft and sensitive clays due to its simplicity, low cost, and suitability for in situ conditions where high-quality sampling is challenging. Despite its extensive application in foundation design and stability analyses, its traditional interpretation is limited to the determination of undrained shear strength. However, there is potential for the test to provide further insights into the soil’s constitutive behavior. Hence, this research explores the use of numerical models validated using experimental results to extract more information from the VST. To do so, we used the Smoothed Particle Hydrodynamics (SPH) method, to model the VST. To validate and calibrate the numerical framework, an experimental program including field VSTs, Cone Penetration Tests (CPT), and triaxial tests was conducted to characterize soil behavior at a local site, and obtain constitutive parameters for the numerical model. The model calibration was then achieved by tuning input numerical parameters to match simulated VST torque–rotation curves with experimental results. Subsequently, a parametric study was conducted to evaluate the influence of a set of constitutive parameters on simulation behavior to extract further information from the test, such as stress-strain relationships with torque and rotation, and in situ lateral earth pressure coefficient (K0). The work contributes to the improvement of the interpretation of VST data and proposes new procedures and correlations to obtain additional constitutive parameters of the soil from the test.

Implementation of an Automated PID Control for Detection of Liquid-Liquid Phase Separation in Droplets

Liquid-liquid phase separation in aerosol particles can influence cloud droplet formation and therefore affect climate predictions. In this work, a relative humidity (RH) control system was optimized to reproduce the conditions of Freedman et al. (2015) and monitor phase changes in organic-inorganic droplets in real time using optical microscopy. A PID-controlled valve regulated the balance of wet and dry air in a sealed chamber. To improve performance, standard bubblers were replaced with fine-pore bubblers for more efficient humidification, and a mobile app interface was developed for easier system tuning and operation. Step-test data were used to model system response and guide PID tuning. The optimized system now provides the stability and reproducibility needed to determine the RH values at which liquid-liquid phase separation occurs.

Performance-Based Design of Shallow Foundations Considering Soil Variability and Climate Change Effects
Traditional geotechnical design relies on deterministic methods and prescriptive safety factors that often fail to account for inherent soil heterogeneity and dynamic environmental changes. This research proposes a shift towards a Performance-Based Design (PBD) framework for strip footings, by integrating stochastic modeling with climate-driven hydro-mechanical coupling. The methodology utilizes Smoothed Particle Hydrodynamics (SPH), an advanced mesh-free numerical approach capable of modeling highly nonlinear behavior and large-deformation processes. The framework is developed in three phases: (1) establishing a deterministic baseline validated against classical analytical benchmarks; (2) incorporating Random Field Theory to quantify spatial soil variability and include it on settlement analysis; and (3), including hydro-mechanical coupling to simulate soil behavior under transient rainfall infiltration. The hydro-mechanical coupling is necessary to evaluate wetting-induced collapse and the degradation of shear strength and stiffness caused by extreme rainfall events. The primary contribution of the work is a comprehensive set of reliability-based design abaci that will enable practicing engineers to directly correlate foundation geometry with specified probabilities of exceeding serviceability limit state thresholds related to settlement, based on stochastic distributions of soil properties and evolving rainfall pattern likelihoods. By explicitly integrating geotechnical uncertainty and climate resilience, this work provides a quantitative tool for risk-informed decision-making, ensuring that shallow foundation systems remain functional and safe under increasingly variable environmental and subsurface conditions.

Building and Documenting a Reproducible Low-Cost Open-Source Optical See-Through AR Headset for Computing Education (Work in Progress)

Students in introductory computer science courses often struggle to understand how algorithms and data structures evolve over time, as traditional flat-screen visualizations require learners to mentally track abstract state changes, limiting spatial and embodied engagement. Augmented reality (AR) offers a promising way to make these processes visible and interactive, yet existing optical see-through head-mounted displays are often too costly for widespread classroom adoption.

To address this barrier, this project focuses on building and systematically documenting a low-cost, open-source optical see-through AR headset for computing education. Inspired by Project North Star, we adapt and refine an existing open-source design with an emphasis on accessibility, reproducibility, and classroom deployment. The system incorporates modular 3D-printed structural components, commodity micro-displays and driver boards, and a repeatable optics calibration workflow based on checkerboard alignment patterns to correct distortion and misalignment.

A working hardware prototype has been constructed, and we will share detailed documentation including a reproducible bill of materials, wiring diagrams, 3D-printable models, and step-by-step assembly and calibration procedures. The calibration workflow includes representative results and troubleshooting guidance to support replication. While interactive AR software is still under development, this work establishes a transparent and replicable foundation that lowers financial and technical barriers for educators seeking to adopt optical see-through AR systems in computing classrooms.

Plexa: Empowered AI Education

AI in education is a hot topic, and for good reason. Professors and Universities are searching for the answers of how they can prepare their students for the future, but the tools available do not provide for their needs. Existing tools fall short in at least one of a few key ways: They are not pedagogically forward, they do not respect data provenance, they are expensive and opaque. Plexa provides educators with the ability to create highly customizable lesson plans for their students to use in a controlled environment, so that they may promote AI literacy and prepare their students for careers in specific disciplines and lives in the age of AI. Chats are restricted to lesson instances; no free form chats. This structure encourages accountable usage, and keeps the focus of the tool on pedagogy. The tool is local first: all data is stored on premise, with the ability to run models within the institution’s network, or outsource compute to any provider of their choice. Plexa will always be fully free, open-source, and permission-less-ly licensed so institutions can examine/modify the code and deploy their own versions without any license related headaches or paying a cent. With this framework, I present a new way forward for what AI education can look like.

Establishing the Effectiveness of the OpenBCI EEG System in Identifying Physiological Markers of Healthy Brains

Parkinson disease (PD) is the second most common neurodegenerative disorder in the United States, affecting 1.1 million people. However, ~20% of PD patients are misdiagnosed because diagnosis often relies on subjective motor assessments by doctors. Electroencephalography (EEG) is a non-invasive, accessible tool that records neural activity using electrodes placed on the scalp, and can be used to improve diagnostics with objective neural biomarkers of PD. The cognitive neuromodulation lab bought the OpenBCI EEG system last spring, and this project sought to identify the reliability of the system. Three experiments were conducted with healthy volunteers. The data for each experiment was preprocessed using EEGLAB to isolate neural data from electrical noise. Then, MATLAB’s signal processing toolbox was used to extract the neural features. The first two experiments looked for robust neurological biomarkers of health data identified with more advanced EEG systems. With subjects alternating between eyes open and eyes closed states in the first experiment and performing thirty trials of finger tapping in the second experiment, two nonmotor biomarkers and one motor biomarker were successfully identified. The third experiment had patients perform finger tapping and spiral drawing, bilaterally, replicating motor tasks in PD assessments. With this data, three biomarkers known to differ between PD and healthy subjects were identified in this healthy cohort, consistent with the data in the literature. Having identified these biomarkers, the reliability of the OpenBCI system is verified and a comparative study between healthy subjects and PD patients will be conducted to identify novel PD biomarkers.

Comparing Functional and Anatomical Division of the Subthalamic Nucleus as a Predictor of Clinical Outcomes in Deep Brain Stimulation

The subthalamic nucleus (STN) is one of the most common targets for deep brain stimulation (DBS), a treatment for alleviating the motor symptoms of Parkinson disease (PD). DBS involves implanting electrodes into the STN to deliver electrical stimulation, with the goal of reducing motor symptoms and improving quality of life for patients with PD. However, its success strongly depends on where within the STN the stimulation occurs. This study aims to determine which segmentation method more closely correlates with motor symptom improvement in DBS.

40 PD patients who previously underwent bilateral STN DBS were included for analysis. Outcomes were measured as percentage improvement across rigidity, tremor, bradykinesia, and overall motor symptom. Anatomical segmentation was performed by dividing each STN into six regions using its center of mass as a reference point. Functional segmentation was derived using the Accolla atlas, which labels the STN into motor, associative, and limbic zones. The atlas was registered to each STN and then the volume of tissue activation relative to the total volume of the STN was calculated to quantify stimulation.

Results showed that functional motor zone activation weakly correlated with rigidity improvement, while other functional zones showed no significant associations. In contrast, anatomical dorsal STN stimulation significantly correlated with rigidity and distinguished responder groups. The dorsal anatomical region demonstrated stronger clinical relevance.