11115 research outputs found
Sort by
AFIT Generative AI Teaching Guidebook
Get PDFAFIT is proud to highlight the Generative AI Teaching Guidebook, a resource designed to provide military educators with practical insights, strategies, and use cases for integrating Generative AI (Gen AI) into their teaching practices. Developed through a collaborative effort involving AFIT faculty across various departments within the Graduate School of Engineering and Management and the School of Systems and Logistics, this digital resource serves as a starting point for educators exploring how to leverage Gen AI in their classrooms. It offers accessible examples and best practices, ensuring utility for instructors of all technical backgrounds. The guidebook provides a comprehensive overview of how Gen AI can enhance education, offering actionable use cases, illustrative examples, and best practices tailored to diverse teaching environments. By addressing both opportunities and challenges—such as ethical considerations and data privacy—it empowers educators to design meaningful learning experiences while fostering discussions about AI\u27s potential and limitations.
The release of this guidebook comes at a key moment as the adoption of Gen AI in education accelerates. It highlights how these tools can be integrated into both traditional academic settings and professional continuing education, such as utilizing Gen AI for educational simulations or modeling in systems engineering. By fostering critical thinking, innovation, and ethical awareness, the Generative AI Teaching Guidebook empowers educators to prepare students for an AI-driven world while advancing AFIT’s defense-focused educational mission
Machine Learning for Reactor Power Monitoring with Limited Labeled Data
Get PDFReal-time reactor power monitoring is critical for a variety of nuclear applications, spanning safety, security, operations, and maintenance. While machine learning methods have shown promise in monitoring reactor power levels, there is limited research on their efficacy in label-starved environments. The goal of this work is to assess the feasibility of classifying nuclear reactor power level using multisource data in scenarios with limited labels. Data were collected using low-resolution multisensors at four nuclear reactor facilities: two large research reactors and two TRIGA reactors. Within each pair, one reactor dataset served as the source and the other as the target in a transfer learning paradigm. Twenty-three supervised models were trained on labeled sequences of magnetic field and acceleration data from each of the target sites. Self-learning and transfer learning methods were applied to the top performing models to assess their classification performance with increasing amounts of labeled data. While reactor power level classification was achieved with a Matthews Correlation Coefficient of up to 0.739 ± 0.003 and 0.622 ± 0.009 with only 400 sequences per power state for the large research reactor and TRIGA target sites, respectively, self-learning and transfer learning leveraging source site data did not improve target classification performance. These findings suggest that alternative methods, such as higher sensitivity sensors, digital twins, or the use of physics-informed models, are required to enable high-performance classification in machine learning approaches to reactor monitoring with a dearth of target ground truth
Radon-Hurwitz Grassmannian Codes
No full textEvery equi-isoclinic tight fusion frame (EITFF) is a type of optimal code in a Grassmannian, consisting of subspaces of a finite-dimensional Hilbert space for which the smallest principal angle between any pair of them is as large as possible. EITFFs yield dictionaries with minimal block coherence and so are ideal for certain types of compressed sensing. By refining classical work of Lemmens and Seidel based on Radon-Hurwitz theory, we fully characterize EITFFs in the special case where the dimension of the subspaces is exactly one-half of that of the ambient space. We moreover show that each such “Radon-Hurwitz EITFF” is highly symmetric, where every even permutation is an automorphism
The Effects of I-127 on a High-granularity LiI(Eu) Bonner Sphere Spectrometer Response Matrix
Get PDFThe measurement of neutron energy spectra using a Bonner Sphere Spectrometer (BSS) requires spectrum unfolding. One necessary component of spectrum unfolding is an accurate neutron response matrix, which describes the energy shift experienced by neutrons when passing through different moderation configurations prior to detection. Leveraging the recently-released ENDF/B-VIII.0 cross-section library, this work details the development of a new LiI(Eu) scintillator response matrix that features improved energy binning granularity as well as increased isotopic and statistical accuracy. Using MCNP6.2, detector responses were modeled at 105 discrete log-equi-spaced incident neutron energies from 1.000 × 10−9 to 2.512 × 101 MeV with a maximum relative error of 1.0%. Comparison with corresponding data from previous work reflects an average variation within 0.025%, reinforcing confidence in the expanded measurements. The impact of 127I within the scintillator crystal and resulting structure within these novel response functions is also explored, along with previously unexplained variations between prior published works
Plane Waves in Non-Symmetric Monoclinic Media
No full textThe propagation of plane waves along the preferential z-axis in non-symmetric monoclinic media is investigated. It is shown, using an eigenanalysis, that the two fundamental polarization states are generally non-orthogonal, although the electric and magnetic field intensities within a given polarization state are orthogonal. It is demonstrated that the two polarization states are orthogonal only if the monoclinic media tensor properties are symmetric. Future work is also discussed, including the analysis of planes waves propagating along a general oblique direction and the electromagnetic characterization of monoclinic media
Fuel Tank Shallow Jet Spurt Characterization and Modeling Using Natural Frequency
No full textAircraft vulnerability reduction efforts aim to increase an aircraft’s ability to withstand damage resulting from the man-made hostile environment. One vulnerability scenario deals with the effects of a round or fragment that pierces a fuel tank wall. Shallow Jet Spurt (SJS) relates to fuel (or any liquid) pulsing out of a fuel tank after taking a direct hit from a ballistic projectile and/or fragment. Modeling and predicting the timing between initial impact and initial spurt is important to vulnerability reduction, as it is tied to initial ignition for fire prediction, modeling, and mitigation. This research seeks to discover a model that correlates SJS timing variables with the natural frequency and other material and structural properties of fuel tank walls. High-speed testing videos and datasets were used to determine a potential relationship between fuel tank structural properties and SJS timing. The initial approach for analysis was to control/standardize all variables to directly correlate this variable of interest with the SJS output. Additionally, a conversion factor was determined to account for liquid-panel interactions while keeping calculations simple. A generalized parabolic correlation between natural frequency and SJS timing was determined, where a larger natural frequency accelerates SJS timing until a critical point. Additional relationships between panel natural frequency, projectile kinetic energy, and SJS timing were also discovered. These relationships require key assumptions that ignore obliquity and first-strike effects, which may affect the exact calculations and require further research but in turn simplify calculations and provides the foundation for further discovery
Data Package Supporting Research on Personalized Learning Path Problem Derived from the Cognitive Theory of Multimedia Learning
Get PDFPersonalized Learning Paths (PLP)s are sequences of learning materials and activities that are designed to deliver personalized education to students. Unfortunately, PLPs are often defined and implemented based on faulty cognitive science practices. The PLP research community may benefit from a problem domain and data set that is derived from a scientifically supported cognitive science discipline. This data is published to support the paper, “A Two-Stage Multi-Objective Personalized Learning Path Problem Based on the Cognitive Theory of Multimedia Learning.” This data set includes the data used in this paper, the questionnaire used to gather learner profile data, and the python code that implements the paper’s algorithms. Researchers are free to use this data to support their own work to advance the state of PLPs
Leveraging Python Interpreters for Concurrency in SeQUeNCe
No full textWith the advent of the Navy Research Laboratory’s announcement of the establishment of the Washington D.C. Metropolitan Quantum Research Consortium (DC-QNet), there has been much interest in the modeling and simulation of the quantum communication network testbed. To that end, we explore in this research the basic functionality of the Simulator of QUantum Network Communication (SeQUeNCe), the developmental Python/C API Interpreters module, and their viability as technologies to be used for high-performance simulation of quantum networks. In this paper, we outline the integration of sub-interpreters with a parallel SeQUeNCe experiment to demonstrate true multi-threading concurrency in quantum network simulations
Adapting for a Local Space Can be Tricky : Designing Units for Teachers to Localize Through Phenomenon Adaptation
No full textLearning science in the context of local phenomena and problems can be powerful for young people. Yet, designing place-based instructional materials is resource intensive, limiting broad access. This study investigates how instructional materials designed for widespread use can support teacher localization through phenomenon adaptation, whereby teachers add or swap phenomena relevant to students\u27 interests, identities, and community. Using design-based research, we developed two upper elementary storyline units and professional learning to support teachers\u27 pedagogical design capacity for phenomenon adaptation. We studied 12 teachers\u27 adaptations during their first implementation of the units by analyzing teachers\u27 interviews, reflections, and professional learning discussions. Findings from both units showed that all teachers added phenomena, with common adaptations including adding student-generated phenomena. In the unit anchored around one phenomenon, teachers extended exploration of existing phenomena, citing student interest and cross-curricular connections as rationale. In the unit motivated by multiple phenomena, teachers added new phenomena to support knowledge building and connect to students\u27 lived experiences. Embedded curricular resources offered low-floor entry points for teachers new to the unit. Supplementary resources showed potential as high-ceiling options for more experienced teachers. Phenomenon adaptation requires teachers to coordinate their knowledge of curriculum, students, and community resources to incorporate meaningful phenomena while maintaining coherence. Challenges included time constraints, high quality of existing materials, limited knowledge of local phenomena, and limited confidence. Implications for curriculum and professional learning are discussed, highlighting the potential to turn curricula designed for widespread use into locally-relevant learning experiences
Investigating Selective Reliability with the Laminar Networking Package
No full textThe guaranteed delivery of all messages in the order sent, is the backbone for the Internet and other networking applications. In time-critical applications, however, this form of reliability hinders performance. In some instances, being able to select different forms of reliability defined by a mix of data delivery attributes is more optimal. To achieve this objective, we will look at the ideals behind the Laminar networking package, which allows a more selective mix of reliability and ordering to be associated with the data. To accomplish this, it defines multiple “channels” between two networked applications