Efficient Performance Recovery Of Pruned Llms For End Devices, Paul Vardhan Bethapudi

Master’s Dissertations

Large Language Models have shown strong performance across reasoning, language understanding, and generation tasks, but their computational and memory requirements make deployment on low resource devices difficult. This dissertation studies efficient performance recovery of pruned LLMs, focusing on whether a pruned model can regain useful task performance through parameter-efficient fine-tuning while preserving the benefits of compression. The work uses Llama-3.2-3B-Instruct as the dense reference model and applies activationaware Wanda pruning at multiple sparsity levels(20%, 30% and 50%). The pruned models are evaluated on language modeling and reasoning tasks using WikiText-2 perplexity and GSM8K (Grade School Math 8000 problems) accuracy. Adaptation …

Investigation Of The Lower Block Rows In The King’S Chamber Of The Great Pyramid Using Ultrasonic Testing With Shear Wave Arrays, Benedikt Maier, Amr G. Hamza, Alejandro Ramirez-Pinero, Khaled Taie, Thomas Schumacher, Olga Popovych, Mohamed Elkarmoty, Mehdi Tayoubi, Hany Helal, Christian U. Grosse

Civil and Environmental Engineering Faculty Publications and Presentations

This study investigates the internal geometry of the granite block rows in the King’s Chamber (KC) of the Great Pyramid of Giza using Non-destructive testing (NDT) with ultrasonic testing (UST) using shear wave echo arrays. As part of the ScanPyramids (SP) project measurement campaigns in 2022, the lower two block rows of all four walls were examined using ultrasonic shear-wave echo arrays (PD8000, Screening Eagle Technologies). The measurement strategy initially included rapid overview acquisitions using an 8-channel UST device to identify potential areas of interest. High-resolution, detailed measurements were then performed at these locations, particularly along the north wall (NW) …

Exploring Gene Regulatory Neural Network Biocomputing Of Bacteria, Adrian Merle Ratwatte

Dissertations and Doctoral Documents, University of Nebraska-Lincoln, 2023–

Artificial Intelligence (AI) has evolved from brain-inspired algorithms into a discipline that increasingly integrates with biological systems. While silicon-based platforms have advanced machine learning, they remain limited in energy efficiency and operation in environments beyond silicon. This motivates biological computing as an alternative, enabling efficient, resource-aware, and reconfigurable computing within living systems. This dissertation addresses these limitations by introducing a bacterial computing framework that models Gene Regulatory Networks (GRNs) as Gene Regulatory Neural Networks (GRNNs). The GRNN mirrors the structure and function of Artificial Neural Networks (ANNs) through gene-gene interactions across trans-omic layers, enabling natural, self-regulating information processing within living …

Magnetic Cellulose Nanocrystal Composites: Synthesis, Properties, Applications, And Opportunities, Mohammad Jahid Hasan, Kishore Chand, Esteban E. Ureña-Benavides, Erick S. Vasquez-Guardado

Chemical and Materials Engineering Faculty Publications

first_page settings Order Article Reprints Open AccessReview

Magnetic Cellulose Nanocrystal Composites: Synthesis, Properties, Applications, and Opportunities

Ion Irradiation Of Microelectromechanical Resonators, Hengky Chandrahalim, David D. Lynes

AFIT Patents

A method of enhancing an electromechanical coupling coefficient of a microelectromechanical (“MEMS”) device. The method includes applying, to a fully fabricated MEMS device, heavy particle ion radiation to the MEMS device at a fluence of at least 1×10¹⁴ cm.⁻². According to other embodiments of the present invention are directed to a bandpass filter comprising a plurality of MEMS devices fabricated in accordance with the methods provided. The MEMS of the plurality are electronically or mechanically coupled.

Improved Method Of Evaluating Gas-Phase Air Cleaners Considering Adsorption And Desorption Mechanisms To And From Internal Test Chamber Surfaces, Casey Douglas Coffland

Dissertations and Theses

Portable air cleaners are playing an increasingly important role in managing indoor air quality in response to pressing issues like wildfire smoke events. Smoke is a complex mixture of gases and particulate matter; while established test methods exist for particulate matter, the dynamic behavior of gaseous pollutants is not well accounted for in current test methods. This is because prevailing air cleaner testing methods typically treat background losses as irreversible, whereas gases can engage in partitioning to and from test chamber surfaces. Ignoring these sorption interactions causes inaccurate estimation of an air cleaner's impact.

A new approach has been developed …

Visionarium, June 2026, Full Issue

International Journal of Emerging and Disruptive Innovation in Education : VISIONARIUM

No abstract provided.

Toward A Metric For Disciplinary Learning In The Age Of Ai: The Unblooms™ Metacognitive Awareness Scale And Discernment Rate In Ai-Mediated Learning, Tina R. Austin

International Journal of Emerging and Disruptive Innovation in Education : VISIONARIUM

Telling students to “reflect on their own thinking” has become insufficient in AI-mediated learning environments. When students are rewarded for polished outputs, cognitive offloading to AI tools becomes rational, and traditional snapshot assessments (single-moment evaluations of task completion) produce false signals about whether durable learning has occurred. This problem is sharpened by the broader shift in AI learning tools toward Socratic tutors, study modes, Khan Academy's Khanmigo, and agentic systems that can shape the learner’s process over time. Lodge and Loble (2026) distinguish beneficial cognitive offloading, which frees working memory for intrinsic learning, from detrimental outsourcing, which bypasses the cognitive …

Metacognition As Disciplinary Infrastructure In Ai-Mediated Learning, Tina R. Austin, Jason Gulya, Nick Potkalitsky

International Journal of Emerging and Disruptive Innovation in Education : VISIONARIUM

Metacognition is often presented as a response to generative AI’s disruption of teaching and learning, yet the term has become too generalized to guide practice. In AI-mediated environments, asking students merely to “reflect on your thinking” is insufficient. Because AI tools can redistribute cognitive labor, their educational value depends on how students use them and whether that use supports disciplinary forms of reasoning. This essay argues that metacognition must be understood as disciplinary infrastructure: students cannot effectively monitor their thinking without understanding the epistemological and ontological demands of the field in which they are working. Classrooms therefore become sites where …

Optimizing Human Capital In Ai-Enabled Architectures: A Systems Constraint And Capability Analysis, Jeremy A. Schlegel, Jennifer Daffinee

International Journal of Emerging and Disruptive Innovation in Education : VISIONARIUM

Artificial intelligence (AI) comprises not only models, but full socio-technical systems involving data pipelines, instrumentation, human-machine interfaces, deployment architectures, and organizational processes for design, monitoring, and evaluation. Using a systems-oriented analytical framework, this paper argues that despite accelerating advances in AI capabilities, human capital remains the enduring and dominant system constraint. Human interfaces define throughput limits in areas such as prompt engineering, data-stream curation, adjudication of model outputs, and the orchestration of hybrid automation workflows including robotics, scraping, and digitization. Synthesizing emerging research across human-AI interaction, machine-learning lifecycle management, organizational adoption, and adult learning theory, we present a socio-technical evaluation …

Law Or Flaw: A Double – Blind Study Comparing Student Comprehension Of Real And Ai Generated Legal Case Briefs, Grant Shostak, Nick Wintz, Melissa A. Petkovsek

International Journal of Emerging and Disruptive Innovation in Education : VISIONARIUM

In recent years, the U.S. legal system has seen an increase of legal filings using fictitious court cases or legal propositions generated by artificial intelligence (AI) (Stokel-Walker, 2026). Rules of professional responsibility require that lawyers review filings in which AI was used to ensure their accuracy (Missouri Bar, Office of Legal Ethics Counsel, 2024, Opinion No. 2024‑11) Despite this mandate, filings with fictitious cases and incorrect statements of law are being filed. Besides posing a threat to the parties to an action, such filings may set an unwarranted precedent for future cases. They also tie up court resources searching for …

A Smart Contract Framework For Project Financing In Gold Mining Operations: Implications For Improved Financial Forecasting, Silvia Lydia Priskilla, Mohammed Ali Berawi, Mustika Sari Dr.

Smart City

Gold mining projects are characterized by high capital intensity and significant operational variability, making financial forecasting particularly sensitive to the quality and timing of cash-flow data. In practice, conventional financing systems rely on multi-stage administrative processes that create delays between work completion and payment realization, reducing the reliability of financial records used for investment evaluation.

This study aims to develop a smart contract-based financing framework to improve the accuracy of profitability projections in gold mining projects. A case study approach is employed using operational and financial data from a mining project in Sumatra over the period 2023–2026. The analysis combines …

Micromechanical Behavior Of Copper Tailings And Sulfide-Bearing Ore, Odai Gifty, Sai Sandeep Chitta

Michigan Tech Publications

The mechanical performance of mining-derived granular materials is increasingly important for their reuse in geotechnical and transportation infrastructure, including the stability of tailings dams, yet their micromechanical crushing and frictional behaviors remains poorly understood. This study investigates single-particle uniaxial strength, failure mechanisms, and inter-particle loading behavior of copper tailings (CT) and an iron-sulfide-rich Ni–Cu-bearing material (FB), alongside a monomineralic quartz sand reference. Single-particle crushing tests at multiple loading rates were used to quantify strength distributions and failure mode frequencies, revealing that splitting governed most failures, particularly for FB, whereas quartz exhibited a broader spectrum of failure modes, including explosive and …

Label-Flip Attack Detection Via Trust-Weighted Aggregation In Federated Learning For Underground Mine Security, Md Sazedur Rahman, Sanjay Madria, Samuel Frimpong

Computer Science Faculty Research & Creative Works

Underground mining operations are increasingly dependent on autonomous vehicles, robotic drilling systems, and intelligent inspection platforms operating in confined, GPS-denied tunnel environments. These systems rely on distributed perception models to interpret navigation cues, hazard warnings, and environmental signals in real time. While centralized deep learning can enhance model performance, transferring raw operational data across mining sites introduces serious confidentiality and security risks. Federated Learning (FL) offers a privacy-preserving alternative by enabling collaborative model training without sharing local datasets. However, deploying FL in underground mining introduces several critical challenges: (i) Training labels may be modified either maliciously by compromised clients or …

Rf-Derived Viscoelastic Modeling For Blood Clot Occlusion Sensing, Abdulrahman Mohammed Aldosari

Thesis/ Dissertation Defenses

Early and reliable detection of vascular thrombus remains a critical challenge for portable and non-invasive diagnostic systems. This work presents an RF-based sensing methodology for blood clot occlusion sensing through the extraction of viscoelastic clot characteristics derived from microwave reflection coefficient (S₁₁) measurements. A compact single-port coaxial probe is employed to interrogate the vessel–tissue system over a wide frequency range (1–25 GHz), where the measured RF response is used to derive equivalent viscoelastic and electrical model parameters associated with clot formation and progression. Both lumped-element RLC and distributed transmission-line representations are utilized to establish the relationship between the measured dielectric …

Integrasi Metodologi Hazop Dalam Pengendalian Risiko Dan Keberlanjutan Operasional Pada Unit Pemulihan Urea: Studi Kasus Pada Industri Pupuk, Riny Yolandha Parapat, Arin Nur'aini Putri, Aryasatya Ramadhan Sukresno

National Journal of Occupational Health and Safety

The fertilizer industry is one of the chemical sectors with high-risk potential due to its operational processes involving hazardous materials as well as extreme pressure and temperature conditions. This study aims to identify and analyze hazards in the urea recovery unit using the Hazard and Operability Study (HAZOP) method, while also formulating effective risk control strategies to support the operational sustainability of the plant. The study was conducted directly at a commercial fertilizer plant in West Java using a semi-quantitative approach, which included field observations, review of technical documents such as Process Flow Diagrams (PFD), Piping and Instrumentation Diagrams (P&ID), …

Discrete Fracture Network Application To Rock Slope Stability In An Open Pit Mine, Elvis Karikari Mensah, Erzah Ackah, Reginald Hammah, Hani Mitri

Journal of Sustainable Mining

The stability of rock slopes in open pit mines is crucial for the safety and efficiency of the mining operation. Conventional stability analysis methods, such as kinematic and limit equilibrium analyses, primarily focus on identifying structural failure mechanisms and evaluating their factors of safety. Although insightful, these approaches do not accurately estimate failure volumes and block locations due to their limited consideration of joint frequency and persistence, which are key parameters in understanding block geometries. Discrete fracture network (DFN) modelling addresses these limitations by explicitly simulating rock mass discontinuities in 3D, which automatically incorporates joint spacing and persistence.

This paper …

A Storm-Ai Primer With An Operationally Representative Example, Shiu-Kai Chin, Matthew Clark, Garrett Katz, Jae C. Oh, William Young

Electrical Engineering and Computer Science - All Scholarship

A STORM-AI Primer with an Operationally Representative Example presents the Systems-Theoretic Technical and Operational Risk Management with Artificial Intelligence (STORM-AI) methodology, a transdisciplinary framework for establishing the trustworthiness of mission systems that may incorporate AI components. The core value proposition of STORM-AI is mission assurance—ensuring that mission systems behave with the predictability and proportionality required by senior leadership and demanded by international humanitarian law. STORM-AI achieves this by faithfully preserving commander's intent from high-level mission descriptions down to hardware-level behavioral specifications through a structured sequence of four formally grounded models: a Mission Model, a Protection (Loss Control) Model, a Security …

Toward Passwordless Document Encryption, Isaac Henry Teuscher

Theses and Dissertations

Document files with sensitive information are used across nearly every industry. In recent years, cyberattacks targeting cloud-based file transfer applications have resulted in millions of sensitive documents being exposed. Although document encryption methods exist, they are often limited in usability, security, or deployability. Password-based encryption, the most widely available option, remains vulnerable to brute-force attacks, unauthorized sharing, and human error. We address these gaps in three phases. First, we present a structured comparative framework adapted from the usability-deployability-security model to evaluate nine document encryption methods across 15 design properties, identifying the benefits and limitations of current approaches. Second, we design …

Hierarchical Parcel Swapping: A Novel Model In Turbulence, Masoomeh Behrang

Theses and Dissertations

Turbulent mixing plays a central role in a wide range of engineering and natural systems, particularly in reactive flows where turbulence--chemistry interactions strongly influence reaction rates, flame structure, and emissions. Accurately modeling these interactions remains a major challenge due to the multiscale and stochastic nature of turbulence, as well as the nonlinear coupling between transport and chemical kinetics. This dissertation develops and evaluates the Hierarchical Parcel Swapping (HiPS) model, a fully stochastic framework designed to represent turbulence-driven transport and scalar mixing without relying on traditional gradient-based closures. HiPS is investigated both as a standalone turbulence model and as a subgrid …

Calibration And Validation Of An Ogden Material Model For Thermoplastic Polyurethane In 3d-Printed Tensegrity Structures, Caleb C. Swain

Theses and Dissertations

This work is motivated by ongoing efforts within the Smart Materials, Adaptronics, and Shock Laboratory (SMASH Lab) at Brigham Young University (BYU) to develop variable-stiffness, external fixators using 3D-printed tensegrity structures. In the process of designing and evaluating these systems, the need arose to accurately model the nonlinear mechanical behavior of thermoplastic polyurethane (TPU), the primary material used in prototype fabrication. This thesis investigates how accurately a calibrated Ogden material model can predict the nonlinear mechanical response of TPU in 3D-printed tensegrity structures. Experimental data were used to calibrate the parameters of the hyperelastic constitutive model through nonlinear optimization. An …

Rheological Protocol To Assess And Optimize Latex Coagulation Dynamics For The Thin Glove Coagulant Dipping Process, Monday U. Okoronkwo, Kok Kong Ng, Wolfram Franke, Gaurav Sant

Chemical and Biochemical Engineering Faculty Research & Creative Works

Many products that directly impact the quality of human life today — gloves, catheters, condoms, and baby bottle teats — are made through the latex-dipping technology. While a variety of methods have been developed – e.g., particle counting, turbidimetry, microscopy, and light scattering – which are suitable for studying the coagulation of latex at very low concentrations, much less work has focused on methods suitable for in-situ characterization of latex coagulation in concentrated solutions (e.g., as relevant to the dipping process). This paper presents a process-relevant rheological protocol for assessing and optimizing latex coagulation dynamics for the thin glove coagulant …

An Access-Control Logic In Hol (Trindemossen 2), Shiu-Kai Chin

Electrical Engineering and Computer Science - All Scholarship

This manual documents the embedding of an access-control logic (ACL) — as presented in the textbook Access Control, Security, and Trust: A Logical Approach (Chin and Older, 2011) — within the Cambridge Higher-Order Logic theorem prover HOL-4. The logic is a multi-agent modal logic with Kripke semantics. Its purpose is to reason about authentication, authorization, integrity, security, and trust. The inference rules of the ACL are proved to be sound. The inference rules are a command-and-control (C2) calculus. The value of the ACL is its support for rigorous, verified, and trustworthy policies for complete mediation, i.e., where all actions are …

Upholding Long-Termism In Basic Research To Shape Leading Advantage Of Automotive Industry, Yubo Lian

Bulletin of Chinese Academy of Sciences (Chinese Version)

General Secretary Xi Jinping has pointed out that basic research is the origin of the entire scientific system and the master switch for all technological issues, emphasizing that it must be advanced with greater efforts and more concrete measures. Upholding long-termism in basic research means placing it at the foundation of the innovation chain and maintaining continuous investment with strategic patience characterized by a long-term horizon and high tolerance for failure. Drawing on the author’s industrial practice, this study systematically reviews the characteristics and roles of basic research throughout the development of China’s new energy vehicle industry. In light of …

Engineering 3d Optical Interconnects And Low-Power Radio Frequency Gaze Tracking, Blake P. Crosby

Theses and Dissertations

This thesis presents two independent research works in optical in- terconnects and radio-frequency (RF)-based gaze tracking. The first work demonstrates a silicon optical waveguide fabricated from four through-silicon vias (TSVs) produced using two-photon-absorption- assisted photo-electrochemical etching. Unlike conventional TSV fabrica- tion methods, this approach enables highly localized, non-line-of-sight etching in silicon using a focused femtosecond laser. Four approximately 5 μm diameter TSVs were etched through a 200 μm silicon wafer in a 7 μm by 7 μm pattern, leaving a central silicon region that functions as a waveguide. The fabricated structure achieved an aspect ratio of approxi- mately 40:1 with …

Analysis On Thermo-Viscous Steady Fluid Motion Over An Impermeable Infinitely Stretched Horizontal Moving Rectangular Surface, P. Krishna, N. Pothanna, K. Spandana

Mansoura Engineering Journal

This study presents analysis on thermo-viscous steady fluid motion over an impermeable infinitely stretched horizontal moving rectangular surface. The numerical results have been found employing R-K method of order 6 shooting techniques developed in Mathematica software ND solve for the flow adaptable equations comprising temperature and velocity. The flow behavior and the impacts of material constraints on the flow region have been analyzed and deliberated taking the help from the generated graphs. The variations of these flow fields have been studied for wide spectrum of physical characteristics which influences the nature of thermo-viscous fluid. The impact of viscosity, constant pressure …

A Conceptual Model Of The External Costs Of Truck Overloading On Road Damage, Emissions, And Traffic Congestion: A Systematic Literature Review, Aleksius Sitepu, Sutanto Soehodho, Andyka Kusuma

Smart City

Freight transport services cannot be separated from the infrastructure and facilities involved, each of which faces its own challenges in delivering optimal performance in the logistics sector. A common challenge in Indonesia’s road sector is the presence of overloaded freight vehicles, which has become a national issue. This phenomenon has a negative impact not only on road infrastructure but also on the environment and traffic. This study aims to develop a conceptual model of the external costs caused by truck overloading practices in terms of road infrastructure damage, increased exhaust emissions, and traffic congestion through a systematic literature review. This …

Leveraging A Centralized Fleet Assignment Management For An Aviation Resilience, Inof Seno Acton Mr., Sutanto Soehodho, Nahry Yusuf

Smart City

Since deregulation in the aviation industry, competition among airlines has intensified. This competition is shown by the increasing number of routes, service times, aircraft, and airports served. This competition causes not all flight services to meet their targets, resulting in aircraft operating less efficiently. Passenger seats are not filled, and flight delays are becoming more frequent. Obviously, this will negatively impact consumers and the airline's finances. The study focused on assigning the fleet used to serve flights to maintain existing aviation services and improve aircraft operational efficiency. This study aims to maximize profits by optimizing fleet assignments through the Centralized …

Decarbonizing Electrical Substations: An Integrative Review Of Green Technologies And Sustainable Practices Across The Infrastructure Life Cycle, Sulthan Syah Ali, Mohammed Ali Berawi, Mustika Sari

Smart City

Decarbonization of electrical infrastructure s becoming increasingly important in efforts to meet climate change mitigation targets, particularly in reducing emissions from substations. However, conventional substations continue to release significant quantities of GHG emissions throughout their life cycle As an example, a typical 500 kV substation emits around 150,000 tCO2e from construction to decommissioning. In addition, the development of green substation concepts remains limited, particularly in Indonesia where standardized approaches are not yet established. In response to the need for more sustainable power infrastructure, this study aims to identify, classify, and review all types of green technologies, practices, and innovations that …

Optimization Of Sea Transportation Services In The Kepulauan Seribu Using The Vehicle Routing Problem (Vrp) Model, Darmadi Darmadi, Sutanto Soehodho, Nahry Nahry

Smart City

The Kepulauan Seribu regency relies heavily on sea transportation for passenger mobility and goods distribution. However, current systems face efficiency challenges, high operational costs, and potential imbalances between demand and service capacity. This study proposes a framework to optimize sea transportation services in the Kepulauan Seribu using the Vehicle Routing Problem (VRP) method, especially the Capacitated Vehicle Routing Problem – Many Single Depot (CVRP–MSD) model with heterogeneous fleets and mixed cargo (passenger and goods). The main objective is to minimize total operating costs, which include fixed costs of using the vessel and variable travel costs, and unmet demand, both passenger …