Modern Security Information and Event Management (SIEM) systems are beginning to use AI tools to help analysts interpret logs and make decisions about those logs. While this can greatly improve efficiency, it also provides new security risks. This project explores whether attackers can manipulate AI-assisted SIEM systems by injecting malicious content into system logs. We investigate multiple injection methods attackers might use, such as filenames, request headers, and services, and measure how well they can influence the AI's reasoning when those logs are retrieved and used as context. Our hypothesis is that carefully crafted log entries can distort severity ratings or response recommendations. To test this, we built an AI-augmented SIEM pipeline and were authorized to simulate a variety of attacks on real infrastructure to ensure accurate measurements of attack success. We measured how often the AI's output was distorted, how much prioritization was shifted, and whether responses would lead to incorrect decision-making. We also evaluated defensive strategies, including log sanitization, modifications to the SIEM pipeline and log aggregator, and structured prompt handling. Our results show that adversarial log content can affect AI-generated recommendations, but by implementing a few mitigation techniques, we can greatly reduce the attack success rate. 

Individually, streak imaging has a well documented history of allowing visualization of very high speed phenomena, and schlieren allows for precise, quantifiable imaging of refractive disturbances within a flow. Here, a combined approach referred to as streak schlieren is implemented by visualizing refractions along a single row of pixels coincident with a background light cutoff and subsequent images are compiled into a single output. The output image appears similar to traditional schlieren taken as a single exposure. This setup is relatively simple and only requires a high speed camera and a light source with a light-to-dark edge called a cutoff which is advantageous for large-scale field experiments. Previous implementations of this technique have provided high quality imagery, but quantification of the magnitude of light refraction has not been done. This work aims to quantify a calibration by mapping pixel intensity through known refraction angles created by a simple lens. This intensity-to-refraction relationship is then applied to experimentally recorded streak schlieren images of supersonic projectiles. Ultimately, the quantified refraction is related to air density through the Gladstone-Dale Law.

Designing an inductive sensing system allows classification of metals using multi-frequency excitation and optimized coil design, intended for usage in recycling and other industries. A transmit coil generates a magnetic field, while a receive coil measures deltas caused by nearby conductive/magnetic materials, captured by ADCs. Digital Signal Processing (DSP) techniques like Fast Fourier Transform (FFT) and In-Phase/Quadrature (I/Q) demodulation extract impedance, phase and amplitude, which are used by a lightweight neural network for classification. Testing showed distinct classification of responses for metals like steel and copper. 

Heat-treated Niobium-Zirconium alloys were investigated for their microstructural evolution and spinodal decomposition across a range of compositions. The phase that these alloys are expected to crystallize into under normal conditions is the body-centered cubic (BCC) crystal structure for the majority of compositions, but the alloy can crystallize in the spinodal structure under aging heat treatments. The alloys were first fabricated by arc-melting samples of the metals together repeatedly to ensure homogeneity. The heat-treatment processing consisted of heating the alloys to 1100 °C for 1.5 hours, followed immediately by a cold brine quench to retain the high-temperature BCC microstructure. This was then followed by an aging treatment at 650 °C for a period of 24 hours in a vacuum, after which the sample was cold-water quenched to maintain the evolved microstructure. A sample from each composition was kept as-cast to serve as an untreated control for changes in physical and crystallographic properties, with a 5 at% Zr composition used to determine a low-Zr concentration baseline. Structural characterization of each alloy was performed via X-Ray Diffraction (XRD) to identify crystal structures and confirm the presence of a spinodal phase, followed by Scanning Electron Microscopy (SEM) to observe the changes in the grains. These results aim to clarify the relationship between composition, thermal processing, and phase evolution in these Nb-Zr systems. 

This work explores how blast loading influences morphology, adhesion, and interfacial behavior of copper plates and a bonded nickel-silver alloy and an explosively etched design. Pre-detonation design parameters are hypothesized to be tunable thus able to produce predictable, desirable surface features. This project aims to establish reproducible relationships between explosion parameters, etching and welding features, and general functional surface performance including resistance to weathering and oxidation. Existing knowledge is incorporated into the process by reviewing and analyzing previous methods and materials used in explosive etching at the Energetic Materials Research and Testing Center (EMRTC). Mechanical testing is performed to assess how explosive welding influences the adhesion and weld strength between copper and nickel-silver alloy. The controlled use of explosion-based processes can enable unique surface finishes, intricate textures, and repeatable designs, creating a foundation for applications that bridge functional performance with aesthetic innovation.

Strain sensors are a fundamental physical sensor employed in the structural health monitoring (SHM) framework due to their ability to accurately quantify in-plane strains on structural surface. There is an increasing need for self-powering capability as supplying an external electrical energy source to the sensor network becomes more challenging with the increasing number of sensors instrumented to cover large structural systems.

In this study, we aim to optimize the design of electrode circuits for stretchable and self-powered mechano-optoelectronic (MO) strain sensing thin films. The MO thin films are known to generate direct current (DC), which varies with in-plane strain, under light. Liquid metal eutectic gallium indium (EGaIn) is used as a top electrode in the MO strain sensor, and the active area for the MO strain sensing is defined by the shape of the EGaIn. To define the liquid EGaIn electrode shape, we employ a mold-injection fabrication to fill the EGaIn in an elastomeric mold with a cavity designed to be a high aspect ratio rectangle. The minimization of active area increases approximation of point strain but limits DC output. Through the MO validation testing – measuring DC output with applied in-plane strain, we optimize the shape of the EGaIn with an improved understanding about how the MO strain sensing characteristic is related to the shape of an MO active area. 

Pneumatic artificial muscles (PAMs) are nonlinear soft robotic devices that produce a variety of forces and displacement paths as a function of pressure, material properties, and geometric parameters. McKibben actuators are a specific type of PAM that are highly researched due to their high force-to-weight ratio and linear displacement path. A McKibben actuator is a multi-input multi-output (MIMO) system, with pressure as the primary controllable input, and force and displacement as either the outputs or disturbance inputs. Transient and steady-state system characteristics were studied based on various input functions. Experimental testing was conducted by applying a variety of voltage inputs (ramp, steady, parabolic) in order to pressurize the McKibben actuator through an electro-pneumatic valve. The force output from the McKibben actuator was measured using digital image processing, and the force data was captured from a tensile testing machine, which was processed in MATLAB. Analyzing a single McKibben control allows for system integration in replicating a human arm by understanding the force output that the actuator will provide. Further work will be done by creating dual-McKibben antagonistic motion to create a control system that has two inputs and two outputs, where choosing force or position is the primary option with an MDoF manipulator. 

Bastnäsite is a rare-earth fluorocarbonate mineral that serves as a natural host for rare earth elements (REE), particularly the light (L)REE cerium (Ce) and lanthanum (La). Understanding the thermodynamic properties of bastnäsite solid solutions is important for improving models of rare earth mineral deposits formation and assessing the mobility of REE in hydrothermal systems. In this study, calorimetry experiments were conducted to investigate the enthalpy of precipitation and mixing behavior of the Ce-La bastnäsite solid solution. The experiments involved mixing aqueous rare earth solutions with varying Ce/La ratios (100/0, 90/10, 75/25, 50/50, 25/75, 10/90, 0/100) and sodium fluoride - sodium carbonate (NaF-Na2CO3) solution within a reaction calorimeter. The enthalpy (heat) associated with mineral precipitation reactions was recorded. Measurement of a range of Ce/La end member and binary solid solution composition permits determining the enthalpy of mixing. After each experiment, reaction products were separated and characterized. Solid precipitates were filtered, dried, and analyzed using Raman spectroscopy and scanning electron microscopy (SEM) to identify mineral phases and evaluate any compositional variations. Remaining aqueous solutions were analyzed using inductively coupled plasma optical emission spectroscopy (ICP-OES) to quantify dissolved rare earth element concentrations and assess partitioning between solid and aqueous phases. The calorimetric results indicate a non-ideal behaviour for the binary solid solution investigated with an excess enthalpy of mixing displaying a symmetric shape. This research is supported by DOE EPSCoR, Basic Energy Sciences, grant DE-SC0026161. 

A Pneumatic Artificial Muscle is a soft robotic actuator that uses compressed air to mimic a biological muscle. A tendon driven muscle uses cable-like materials to produce motion. Hybrid pneumatic-tendon powered muscles combine the pneumatic and tendon driven systems aimed at achieving precise movement, high speeds, and high dexterity. Similar dual-driven actuators are an emerging research area due to the complexity of its design and control. This soft robotic elastic digit is inspired by an octopus arm: using biomimicry to replicate how an octopus solves problems and interacts with objects by maneuvering its tentacles. To characterize the behavior of the proposed soft robotic muscle design, three parameters are studied. The material characterization of 3D printed TPU (Thermoplastic Polyurethane filament), and Silicone Elastic 50A are analyzed and compared using MATLAB. Elasticity, durability, and strength are evaluated to refine alternative actuator designs. Pressurization and tendon displacement are varied to understand how motion is affected and to understand the limitations of a hybrid pneumatic-tendon system. These three parameters help to develop a design methodology for further refinement for specific soft robotic applications. This bio-inspired approach to a soft robotic digit yields the capabilities an octopus arm may provide, including a digit with multiple degrees of freedom. 

Digital twin technology has emerged as an effective approach for monitoring, evaluating, and optimizing complex engineering systems by maintaining a continuously updated virtual representation of a physical asset. In aerospace applications, this capability improves system understanding, operational awareness, and decision making through continuous integration of real time sensor data. Unmanned Aerial Vehicles (UAVs) operate in highly dynamic environments where flight stability and mission reliability depend on both internal system conditions and external environmental factors. Because UAV flight dynamics are nonlinear and strongly influenced by environmental conditions, relying solely on physical testing is often insufficient for comprehensive operational analysis. This study presents a real time digital twin framework designed for monitoring UAV operations and detecting abnormal system behavior using multi sensor data. The proposed system integrates telemetry and environmental measurements including GPS position, barometric pressure, temperature, humidity, carbon dioxide concentration, and particulate dust levels. These data streams are transmitted to a Unity based three dimensional visualization environment, where a virtual representation of the UAV is synchronized with the physical system, enabling continuous visualization of flight dynamics and environmental conditions. Machine learning based anomaly detection models analyze the incoming data streams to identify patterns that may indicate system instability, sensor malfunction, or potential failure conditions. The framework demonstrates how digital twin technology combined with environmental sensing and machine learning can support safer UAV operation, improve monitoring capabilities, and enable future development of predictive maintenance and autonomous aerial system management. 

My SRS project is going to be about NMT’s scholarship system and its superfluous websites. Scholarships are often the lifeblood of a student studying at college, so we should make it as easy as possible to go through the red tape and processes that allow students to access grants and scholarships. The system currently employed by NMT currently has several issues, the main ones being that the websites are bloated, sometimes out of date information with multiple websites telling you the same thing, unorganized and unnecessary, and the other main issue being a lack of accountability by the financial aid department and the difficulty it is to navigate the higher branches of the NMT system. I propose that we condense the multiple websites NMT has with scholarship information down to a single database, that uses filters and keywords to allow students to search for scholarships they can apply for. For better accountability, I propose we add a diagram explaining the “corporate” ladder of each department, so if someone isn’t doing their job correctly or if you have further questions, you know who to take them to. 

Lightning imaging represents one of the most valuable forms of ground-truth data for lightning research, and in the summer of 2025, the first LIGER (Lightning Imaging for Ground-truth and Electric-field Recorder) prototype was deployed at the Workman building. We analyzed parameters such as Flash Multiplicity and Flashes with multi-grounded strike points. The results show that each flash has an average of 2.6 return strokes and 1.4 connections to the ground. Around 28.8 % of flashes contained a single stroke. The data collected can be used to assess the accuracy of space- and ground-based lightning detections, including lightning type classification and detection efficiency. The initial deployment of LIGER was conducted entirely indoors, which significantly constrained the system’s field of view and limited the range of camera orientations. To address these limitations, an outdoor enclosure was developed to enable more flexible operation. A pan-and-tilt mechanism, driven by servo motors, was designed to provide precise control over the camera’s pointing direction. Custom 3D-printed mounts were fabricated to securely house the servo motors and camera assembly. The servos are controlled by an Arduino board, which receives commands from the same compact computer that runs the camera software. Both the Arduino and the servo–camera assembly were mounted within a portable plastic enclosure. The enclosure lid was modified to incorporate a transparent acrylic dome, providing an expanded, unobstructed field of view while maintaining environmental protection. 

Broadband optical and electromagnetic emissions produced during lightning discharges offer insight into the fundamental physical processes involved. The Dragonfly’s Eye Viewing Lightning (DrEVL) instrument was designed to record multi-band optical emissions across a spectrum from 300 nm to 1150 nm. DrEVL was deployed during the Summer 2025 field campaign at Langmuir Laboratory. We also deployed electric field sensors and middle-speed and high-speed cameras for context. In addition, we used data from the Earth Networks Total Lightning Network (ENTLN) and the Geostationary Lightning Mapper (GLM). In this study, we analyze six distinct lightning flashes observed in the field of view of our instruments: two intracloud (IC), two negative cloud-to-ground (–CG), and two positive cloud-to-ground (+CG) flashes, collectively producing 50 individual pulses. For each pulse, both the optical peak intensity and a Channel Detection Score (CDS), defined as the signal-to-noise ratio relative to the pre-event baseline, were calculated and subsequently compared with the corresponding EN peak current measurements. The results reveal that the spectral response varies significantly across different lightning types. Near-infrared channels demonstrated the highest sensitivity, and shorter-wavelength channels (ultraviolet) exhibited reduced sensitivity. While most channels showed a moderate correlation (r ≈ 0.55) between optical peak intensity and peak current, the 1150 nm channel displayed a weak negative correlation (r ≈ −0.24). These findings suggest that specific wavelength bands are more sensitive to particular discharge processes and underscore significant variation in spectral response by lightning type, supporting the development of future classification methods that integrate optical and electromagnetic observations. 

The attenuation of solar irradiance by cloud cover is a highly variable component of the local climate energy budget. Traditional meteorological stations measure this directly, but this project investigates an inverse methodology: utilizing existing photovoltaic (PV) solar panels as localized meteorological sensors. It is hypothesized that local cloud presence and solar irradiance can be accurately predicted at a specific location and time purely through the electrical metrics (specifically voltage) generated by a solar panel. To test this, a pre-compiled, ready-made dataset of PV electrical output will be analyzed alongside historical ground-truth meteorological measurements for the corresponding times and locations. The methodology employs a dual-model machine learning approach: a logistic regression model to classify cloud presence (e.g., cloudy versus clear skies) and a continuous regression model to predict the exact solar irradiance values. By comparing the model predictions to the actual meteorological data, this project aims to demonstrate a strong predictive correlation, establishing whether basic PV voltage metrics can serve as a reliable, measure for local atmospheric observation. 

This study investigates the behavior of low-temperature, low-pressure argon plasma as a 2D axisymmetric fluid subjected to an alternating voltage with various parameters. The primary purpose of this research is to optimize and control plasma-based manufacturing processes. Because computer chips can melt under high heat, understanding the effects of low-temperature plasma is necessary for delicate etching. These low-temperature plasmas have room-temperature neutral particles, but most of their energy is stored in the high-temperature electrons. To model the plasma, we discretized the drift-diffusion fluid equation using the finite volume method. We model our region as an axisymmetric cylinder geometry to simplify calculations. We also enforce that the right boundary condition is grounded, whereas the left boundary condition uses a sinusoidal voltage alternating at 13.56 MHz – a common radio frequency for plasma. We verify our model's outputs by comparing the results to a particle-in-cell collision method by Wilczek (et al.), 2020 and a collisional-radiative fluid model by Tsagkardis (et al.), 2025. After verification, we test and discuss various parameters such as maximum voltage (10 - 500 V), pressure (1.00 - 10.00 Torr), electrode gap length (0.25 - 5.00cm), and frequency (6.78 - 27.12 MHz). Results indicate that plasma characteristics are most sensitive to voltage and pressure changes, while changes in gap length and frequency have minor influences. Future work may expand this model by adopting an implicit approach, allowing for larger time steps and smaller pressure.  

The New Mexico Tech Space Weather Explorer CubeSat project addresses the challenge of designing a reliable, cost-effective 6U cubesat capable of withstanding launch loads and operating in the harsh low-Earth orbit (LEO) environment to study the propagation of VLF whistler waves in the Earth’s magnetosphere and ionosphere. A significant engineering challenge is fitting the different subsystems into the 6U (100 mm by 224 mm by 366 mm) CubeSat frame while simultaneously addressing the mission goals and the launch provider requirements. Independently designing the components can lead to conflicts in the CubeSat, preventing all necessary parts from fitting within the confined space, and part incompatibilities. To address this problem, we used CAD modeling and 3D printing to model the assembly of the components inside the frame. This allows us to validate the fit and the eventual assembly of the satellite. 

Electronic side-channel attacks exploit information leaked by electronic systems, such as power consumption, timing, electromagnetic emissions, and acoustic emissions. As these signals are often noisy and indirectly related to secret data, cryptographic key recovery requires statistical inference. In this work, we use Correlation Power Analysis (CPA) with a Hamming-weight leakage model to recover an AES-128 ECB key from two embedded processors: the ATXmega128 (8-bit AVR) and the Raspberry Pi RP2040 (32-bit ARM Cortex-M0+). Our hypothesis is that the RP2040's increased architectural complexity will require more power traces and produce weaker power correlation compared to the ATXmega128. Both targets run bare-metal to isolate cryptographic leakage from OS-level noise. We compare the two architectures using trace count to successful key recovery, correct-key rank progression, and average correlation separation between correct and incorrect key guesses. Preliminary results on the ATXmega128 demonstrates full key recovery and RP2040 analysis is ongoing. These findings contribute to understanding how processor complexity affects information leakage in embedded systems. 

Single-photon emission is a crucial step for quantum technologies. G-center defects, a well-studied defect in silicon lattices, have shown promise in single-photon emission around the telecom wavelength of 1280 nm. To leverage the advantages of G-centers we need a resonant cavity in silicon with photonic crystals providing these resonant cavities. Because of the sub micron size of these crystals they are hard to fabricate to exact specifications. In this work, we present a silicon photonic crystal device whose mode is widely tunable with piezoelectric actuators. This allows for higher variability in the fabrication process and will allow us to use a photonic crystal whose mode frequency can be made to overlap that of G-center emission. We demonstrate a tunability of ~0.62 nm per volt applied with a device whose theoretical quality factor is ~10^6. 

Spectroscopy proves to be a powerful diagnostic tool for probing transient plasmas to examine their physical properties, temperature, composition, and energy transfer processes from their characteristics emission and absorption signatures. For lightning, it can help quantitatively determine its chemical and energetic impacts on the atmosphere. As such, we have built a code to produce a synthetic spectra for lightning return strokes to compare with experimental data and determine what processes are present. The range of wavelengths spans from ultraviolet into the infrared, including the visible portion of the spectrum (10-1000 nm). It currently models emission and absorption processes for atomic species of oxygen and nitrogen. In terms of emission, recombination and Bremsstrahlung are the primary contributors to the continuum of the spectrum. Model results show that the recombination of oxygen and nitrogen atoms is by far the greatest contributor to the emission budget. For typical lightning return stroke temperatures of 10-30 kK, we see that recombination dominates at the shortest wavelengths (~10-100 nm), with line emissions usually dominating the rest of the vacuum ultraviolet band (~100-200 nm). Leaving all of the visible and near infrared having contributions from Bremsstrahlung and line emissions, with lines seeming to become more prevalent with increasing wavelength. Future work will include contributions from more molecular species, and involve using this model to estimate plasma temperature, chemical yield, and energetic impacts of lightning in Earth’s atmosphere. 

Mechanoluminescent materials as self-powered strain sensors have been a matter of interest for researchers for a long time now. The sensors work by applying strain on the matrix which induces a piezoelectric field in the embedded phosphors that release electrons, which emit visible light. The spectral characteristics of the emitted lights can reveal the type and intensity of the mechanical stimuli acting on the material. However, obtaining said spectral data can be costly and time consuming as spectroradiometers are quite expensive and the measurement requires a proper laboratory setup, which is not practical in practical use cases. Furthermore, the emitted light may not be bright enough for the spectroradiometers to detect properly. Hyperspectral regeneration is a method where the 3 channels of an RGB image can be subdivided into multiple channels that all can have intensity data for different wavelengths and can produce a similar spectrograph like data from RGB images. This project uses hyperspectral reconstruction of RGB images taken from a high-speed camera to generate the spectral information from the light emitted from the stress sensors and observe how the response changes with gradual change in strain. 

We present an analysis of the physical attributes of the CH CYG binary star system. We obtained data of the system in the K and H bands using the Center for High Angular Resolution Astronomy (CHARA) Array. Using this data, we tested different modeling techniques using Python and Julia to determine the best one for the star. We tested limb darkening, uniform disk, and elliptical fitting. With these results, we determined the size of the star in solar radii. We then used spot fitting to gain an understanding of the average convection cell size on CH CYG, and tested this against theoretical cell sizes calculated using the known physical parameters of the star. 

The NMTSWE satellite will orbit the Earth at an altitude of 500 km, collecting space weather data, transmitting a beacon signal to an array of ground stations, and downlinking collected data to receiver ground stations. At the same time, solar panels must be pointed optimally toward the sun to collect the power that the satellite needs to operate. To fulfill the mission goals, optimally, the satellite orientation (called attitude) must be controlled. A common way to control the attitude of small low-Earth orbit satellites is with magnetic torque coils. As a current passes through a coil, a torque is applied because of the interaction with the Earth’s magnetic field. Satellites usually have three mutually orthogonal torque coils which can be energized depending on the required torque and orientation of the magnetic field. The challenge is to maintain the required attitude as the satellite orbits the Earth while the magnetic field direction and strength is changing in the satellite frame. In the presentation I will discuss the attitude requirement as well as some simulations of magnetic torquing. The goal is to develop torquing algorithms which accurately and power-efficiently control the attitude of the satellite in all parts of its orbit. 

The detailed evolutionary sequence of high-mass (M ≥ 8 𝑀 ) star formation remains unknown. It is important to study the evolution of these stellar objects as their formation process strongly influences their surroundings and plays a major role in shaping the interstellar medium. One useful way to investigate this process is by analyzing the morphology of high-mass star-forming regions that contain protostellar objects in the early stages of their evolution. In this research project, we present Very Large Array (VLA) continuum observations and 1.3mm Submillimeter Array (SMA) continuum observations toward the high-mass star-forming region ISOSS J23053+5953. Within this region lies SMM1, a massive core with evidence of embedded young protostellar objects. By overlaying SMA contours on the VLA data, we analyze the morphology of the region and investigate the structure, shape, and spatial distribution of SMM1, while also aiding future interpretation of spectral line emission data from the same region. From SMA continuum data, we are able to determine the total mass of the core as well as create spectral energy distribution plots (SEDs) when incorporating the VLA continuum data. We found that from the SED plot relating to flux density and frequency is consistent with emission from an ionized jet. From these results and upon further investigation, we hope to further our insight into the process of high-mass star formation.

Human-robot collaboration (HRC) in modern manufacturing continues to grow. Most of the current shared autonomy systems employ non-adaptive control techniques. These non-adaptive control techniques are based on fixed robot behaviors, and thus do not adapt robot actions based on environmental risk or operator's cognitive state. These approaches enhance operator situational awareness and operators are required to make decisions regardless of their mental workload. Therefore, while nearly all of the existing HRC systems are risk-aware, very few have incorporated operator's cognitive status into the autonomous control of robots. This proposal develops a Cognitive Aware Adaptive Autonomy (CAAA) framework that dynamically regulates robot autonomy by integrating both the predicted task risk and the cognitive readiness of the operator. The CAAA framework includes two primary elements. First, an XR-enabled digital twin simulates the collaborative work space, and generates a task risk score (R). Second, a cognitive monitoring module measures operator workload and attention using various behavioral metrics to generate a cognitive score (C). Finally, a decision module integrates R and C to dynamically regulate autonomy of the robot and select one of four operating modes: manual assist, shared control, robot lead, and safety restriction. The proposed system will be tested in a precision assembly study with a 6 degree-of-freedom robotic arm working under disturbances, while also varying the operator's cognitive load. The expected benefits of this system include improved safety, increased precision, better workload balance, and greater operator confidence.