Independent Research Projects

Title: Integration of Data Acquisition and Radiation Detection

MIDN: Jadiel I. Arana

Faculty Advisors: Dr. Hau Ngo, Dr. Brian Jenkins, Dr. Hatem ElBidweihy

Abstract:

This research paper presents a novel approach to portable radiation detection through the integration of field programmable gate arrays (FPGAs) and system-on-chips (SoCs). Traditional radiation detectors, often reliant on high-voltage photomultiplier tubes (PMTs), face limitations in portability and power efficiency. By employing Silicon Photomultipliers (SiPMs) with significantly lower biasing voltages, this study develops a compact detector design that enhances operational flexibility. The proposed system utilizes a Terasic ADC-SoC FPGA Development board, integrating data acquisition and real-time signal processing capabilities into a unified platform. Experimental results demonstrate that the FPGA + SoC architecture provides substantial improvements in data acquisition speed and efficiency compared to mixed-signal setups and an ability for fast real time processing with a high accuracy for determining the presence of a radiation source. Furthermore, the paper introduces a method for noise floor analysis, improving calibration processes and minimizing the impact of dark pulses. This work underscores the potential of FPGA + SoC based radiation detectors to meet the growing demands for portable and precise detection systems across various industries, including environmental monitoring and nuclear safety. The findings indicate that such systems can be tailored for specific applications, providing low-cost and efficient solutions for real-time radiation monitoring in diverse environments.

Poster

Title: Machine Learning Algorithms for Neutron and Gamma Radiation Discrimination

MIDN: Abigal Artman

Faculty Advisors: CDR Jennie Hill, Dr. Latta, Dr. ElBidweihy

Abstract:

The uncontrolled use of radioactive materials, such as Special Nuclear Material (SNM), is a long standing issue for national security and nuclear non-proliferation applications. Current detection methods for radioactive materials are riddled with inefficiencies and are in need of advancements to match current threats. Machine learning algorithms offer improvements in the discrimination of neutron and gamma radiation, especially for the challenges of in-situ applications. This research explores machine learning classifiers for neutron-gamma radiation discrimination, focusing on achieving high sensitivity and low false positives with minimal training data for in-situ applications. Using a dataset of one million events, we trained and evaluated 34 classifiers using MATLAB’s Classification Learner App. Each event was recorded as a voltage pulse generated by an EJ-309 scintillator coupled with a photomultiplier tube. Four primary features were extracted for analysis, with time-of-flight (ToF) data included as a fifth feature in a secondary evaluation. Through a systematic process of training with increasingly less data, each models’ performance was evaluated by F1 score and sensitivity. Results demonstrate that classifiers can achieve high performance with as few as 80 events, with F1 scores and sensitivity values within 1% of those achieved with a 10,000-event dataset. While pulse shape parameter (PSP) was identified as the most critical feature, classifiers trained with all features except PSP still demonstrated strong performance, showcasing their resilience even when deprived of the most separable feature. ToF data provided minor improvements across all dataset sizes but was largely unnecessary for achieving high classification accuracy. This research emphasizes the effectiveness of specific machine learning classifiers and the versatility of feature combinations for reliable neutron-gamma discrimination in both field and laboratory settings.

Poster 

Title: Enhancing Code Quality with Generative AI: Boosting Developer Warning Compliance

MIDN: Hansen Chang

Faculty Advisors: Dr. Christian DeLozier

Abstract:

Programmers have long ignored warnings, especially those generated by static analysis tools, due to the potential for false-positives. In some cases, warnings may be indicative of larger issues, but programmers may not understand how a seemingly unimportant warning can grow into a vulnerability. Because these messages tend to be long and confusing, programmers tend to ignore them if they do not cause readily identifiable issues. Large language models can simplify these warnings, explain the gravity of important warnings, and suggest potential fixes to increase developer compliance with fixing warnings.

 Poster

Title: Fiber Optic Sensing for Hardware Anomaly Detection

MIDN: Alex Conway

Faculty Advisors: Dr. Brian Jenkins, Dr. Hau Ngo, Dr. Dan Opila

Abstract:

Fiber optic sensing is being used to observe hardware anomalies on two categories of semiconductor components. Distributed fiber optic sensors were first attached to a programmable development board. The sensors monitored temperature during specific operations commonly performed on programmable hardware platforms such as a field- programmable gate array (FPGA). The sensors detected temperature changes that were consistent for identifying specific operations. The fiber optic sensors were also used to measure semiconductor package temperatures on power diodes and provided intrinsic high voltage isolation and low electromagnetic susceptibility. Thermal characterization during switching or power cycling is important for monitoring device behavior in power electronic systems. Ultimately, the goal of this project is to prove that fiber optic sensing may be a viable technique to diagnose hardware anomalies that result during specific computing functions, such as malware, or for detecting thermal changes on semiconductors embedded in power module systems.

Poster

Title: Fabrication of Capacitive Sensors

MIDN: Yvonne Fu

Faculty Advisors: Dr. Hatem ElBidweihy, Dr. Connor S. Smith

Abstract:

Interdigitated capacitors (IDCs) are planar and used for its ability to be more easily fabricated, lower power consumption, and faster response time. Changes in dimensions and interference in its electromagnetic field can be indicated by its capacitance. Increasing the device’s characteristic capacitance improves its sensitivity and precision when used in environmental sensing applications. Design 3 with 1-mm wide fingers and Design 4 with 0.2-mm wide fingers were fabricated on a Cu-FR4 substrate using maskless photolithography and etching. The measured average capacitance for Design 3 was 45.59 pF while Design 4’s average measured capacitance was 12.92 pF. Design 4 was fabricated using three different brands, with the Midwest Circuit Technology substrate exhibiting the highest average capacitance of 15.806 pF. Simulations predicted twice the capacitance on Cu-Al2O3 and Cu-AlN substrates, as compared to Cu-FR4, but a successful print was not obtained on the Cu-Al2O3 substrate due to over-etching. Challenges were found when gaining familiarization with the devices and the etching process.

Poster

Title: Additively manufactured Frequency Reconfigurable Patch Antennas Using Dielectric Millifluidics

MIDN: Colby Grosse

Faculty Advisor: Dr. Connor S. Smith

Abstract: 

As the communication and frequency band has opened up having multifunctional patch antennas will allow communication devices to use more frequencies without needing more space. Frequency reconfigurable patch antennas are a subset of multifunctional patch antennas that change their resonate frequency. Using one of many methods they are able to shift their resonate frequency by changing their characteristics. The reconfigurable patch antennas in this report use dielectric fluid in a tunnel below the leading edge of the patch antenna. Different levels of fluid change the characteristics of the antennas and subsequently the resonate frequency. The antenna was created using a “Cricut” crafting CNC. Using the free software and a section of copper tape the antenna and feedlines were cut out. The substrate was modeled in “Solidworks” and upload to the FormLabs Preform software. A Formlabs Form 4 stereolithography printer with ClearV5 resin was used. The antenna is characterized with a Vector Network Analyzer that measures the difference in sent and reflected energy. With everything set up the dielectric fluid was input into the chamber at varying levels. First the edge of the antenna, then the middle, the far edge and finally full tunnel. The VNA collected the reflections and the data was input into Excel and graphed. The data showed the different amounts of dielectric fluid changed the resonate frequency of the antenna. This project was a success. The use of a dielectric fluid can change the properties of an antenna. This technique can be used to help produce more robust antennas for many applications but mainly in communication.

Poster

Title: Data-Driven Approaches for Detection of Lidar-Readable Barcodes in Autonomous Vehicles with Machine Learning

MIDN: Tony Harrington

Faculty Advisors: Dr. Kevin Galloway

Abstract: 

Lidar is a powerful tool commonly used in self-driving cars alongside cameras for sensing of the environment. Image recognition can be costly in terms of computing, so methods have been studied to use the already incoming lidar data to receive additional information. Lidar readable barcodes have been investigated to encode information for these vehicles. Lidar intensity sensors have produced accurate readings, but the decoding system is first required to detect the presence and position of the barcode in the environment. A model-based approach is currently used for barcode detection, but machine learning methods can be utilized to develop a data-driven approach. The primary learning algorithm was fine trees in a regression configuration. Training and testing data sets were produced in the lab, and the machine learning model was trained, tested and tuned in an interactive manner. The resulting detection model was reasonably successful despite apparent overfitting, and the process provided insights for future improvements in barcode detection.

 Poster

Title: Preventing Client-Side Exploits in Games with Capability Architectures

MIDN: Joe Oster

Faculty Advisors: Dr. Christian DeLozier

Abstract: 

Client-side exploits, such as wall-hacks and aimbots, affect the competitive integrity of games. Modern competitive games prevent such exploits through kernel-level monitoring that not only invades the player’s privacy but also potentially opens a channel for kernel-level attacks on the player’s computer. Capability architectures, which attach additional metadata to memory, offer a solution to preventing client-side exploits that can be palatable to both players and game designers. By compartmentalizing memory with restricted permissions, games can prevent players from viewing or modifying memory associated with the game directly. However, games need not have access to the entire system at the kernel-level to perform this task. Language extensions concerning memory are employed to assign permissions to specific areas of memory, ensuring that data is only accessible to functions with the appropriate authorization. This paper focuses on using memory compartmentalization facilitated by a language extension capability architecture to prevent aimbots from accessing restricted data. Our language extension architecture leverages C++ smart pointers to compartmentalize data and allow permissions on specific functions. These permissions allow us to have a restricted implementation of an aimbot detector. The capability architecture is implemented in a sandboxed first-person shooter. Our imple- mentation will then be evaluated against an outdated hardware solution that uses an enclave to secure an aimbot detector.

Poster

Title: Drone-Based Method for Field Deployable RF Location Finding

MIDN: Myla Reardon

Faculty Advisor: Dr. Gregory Coxson, Dr. Connor S. Smith

Abstract: 

Radio Frequency (RF) Location Finding is the collection and processing of transmitted radio signals to determine the location of their origin. Traditional RF Location Finding typically uses two or more receiving (Rx) antennas and requires knowledge of their precise locations and relative angles to determine the origin of a RF signal with a reasonable degree of accuracy. This project employs a single, drone- mounted Rx antenna and a simple distance and signal strength equation to predict the distance to the origin of an RF signal. A stationary signal generator, emitting a signal using a dipole transmitting (Tx) antenna, and a mobile spectrum analyzer, with a similar dipole Rx antenna, are mounted on carts for ease of transportation. The spectrum analyzer moves in a straight line perpendicular to the signal’s path, with signal strength measurements taken at even intervals along the path. The collected signal strength data is entered into Excel, and used with the signal strength and distance equation to approximate the closest distance between the transmitter and the receiver. The most recent series of data collection generally demonstrates the predicted Gaussian distribution for signal strength along the perpendicular path, but fails to approximate the value for the closest distance with a reasonable degree of accuracy. Potential limitations to this technique include variable Tx signal strength and the fact that even small discrepancies in the transmission path result in a significant amount of interference which distorts the signal strength pattern at the receiver. The benefits of this method for RF location finding over traditional methods include the necessity for less hardware with only one required receiver, and its increased mobility which can be crucial in a dynamic and unpredictable environment.

Poster

• Integrated Cooling of Additively Manufactured Optical Components with Applications in Directed Energy
• Computational Sprinting on Mobile Devices with Machine Learning
• Aerosol Jet® Printing of Dielectric Elastomer Actuators Using Flexible Dielectric and Conductive Inks
• Liquid Metal Antenna
• Vitrovac/PZT Magnetoelectric Laminate Composites for Wireless Power Transfer for In-Vivo Biomedical Devices
• Additive Manufacturing of Optical Cloaking Components
• Correlating Ransomware Functionality to Hardware Performance Counters
• Block the Bot: Implementing Aimbot Detection Using Intel SGX in Open Arena
• Utilizing Wavefront Sensors to Optimize and Experimentally Analyze 3D Printed Optical Components and Systems