26800 research outputs found
Sort by
Evaluation of Cyberattack Impacts on Distribution Grid Networks
No full textThis paper presents a robust algorithm to enhance the resilience of electrical distribution grids against cybersecurity threats like malware and DDoS attacks that compromise metering integrity in advanced distribution systems. The solution utilizes distributed metering across DERs, microgrids, and ADMS/DERMS/SCADA systems to improve data reliability and control. A key feature is the dynamic trust assignment using parameters α, β, and γ to evaluate data dependability and consistency. By identifying outliers and anomaly patterns, the algorithm detects security events and faults, adapting trust levels to mitigate disruptions, including DDoS attacks on smart meters and communication links. The results presented in the case study highlight the efficacy of the proposed trust-based framework that addresses key cybersecurity challenges in the distribution networks
You Don\u27t Need All Attentions: Distributed Dynamic Fine-Tuning for Foundation Models
No full textFine-tuning plays a crucial role in adapting models to downstream tasks with minimal training efforts. However, the rapidly increasing size of foundation models poses a daunting challenge for accommodating foundation model fine-tuning in most commercial devices, which often have limited memory bandwidth. Techniques like model sharding and tensor parallelism address this issue by distributing computation across multiple devices to meet memory requirements. Nevertheless, these methods do not fully leverage their foundation nature in facilitating the fine-tuning process, resulting in high computational costs and imbalanced workloads. We introduce a novel Distributed Dynamic Fine-Tuning (D2FT) framework that strategically orchestrates operations across attention modules based on our observation that not all attention modules are necessary for forward and backward propagation in fine-tuning foundation models. Through three innovative selection strategies, D2FT significantly reduces the computational workload required for fine-tuning foundation models. Furthermore, D2FT addresses workload imbalances in distributed computing environments by optimizing these selection strategies via multiple knapsack optimization. Our experimental results demonstrate that the proposed D2FT framework reduces the training computational costs by 40% and training communication costs by 50% with only 1% to 2% accuracy drops on the CIFAR-10, CIFAR-100, and Stanford Cars datasets. Moreover, the results show that D2FT can be effectively extended to recent LoRA, a state-of-the-art parameter-efficient fine-tuning technique. By reducing 40% computational cost or 50% communication cost, D2FT LoRA top-1 accuracy only drops 4% to 6% on Stanford Cars dataset. The extended version of this paper can be found in http://arxiv.org/abs/2504.12471
Additive Manufacturing of Ti-Based Alloys: Microstructure, Mechanical Properties, and Corrosion Behavior
No full textImplementation of rapid prototyping additive manufacturing (AM) has revolutionized and accelerated applications of materials in manufacturing aviation components, personalized prosthetics, and implants with complex geometries and microstructures. Due to its highly-versatile material processing and precise structure control, AM has been widely employed in manufacturing titanium (Ti) and its alloys, as seen in AM-based techniques, including fused deposition modeling (FDM), techniques based on powder bed fusion (PBF) including selective laser sintering (SLS), selective laser melting (SLM), and electron beam melting (EBM), and methods based on direct energy deposition (DED) such as laser engineered net shaping (LENS) and wire arc additive manufacturing (WAAM). Processing Ti alloys, however, is a challenging task due to its complications associated with the processing parameters and the influences on structures and properties of manufactured parts or products. In this paper, we review the AM of Ti and its alloys, focusing on the microstructure of AMed Ti-based parts or products and their mechanical and corrosion properties. This study also addresses the potential of AM methods for the production of complicated components, including cellular structures, and their utilization in the aviation and medical fields. Key challenges and trends of the FDM, PBF-based, and DED methods are also identified and discussed, along with recommendations for future studies on AM-fabricated Ti alloys for improved properties
BRIDGING INDIGENOUS AND WESTERN KNOWLEDGE SYSTEMS: CENTERING OJIBWE PERSPECTIVES FOR JUST FOREST-CLIMATE FUTURES
No full textThis dissertation addresses critical gaps in how environmental science and policy engage with Indigenous knowledges that hinder just and sustainable environmental futures. Through a multi-year partnership with the Keweenaw Bay Indian Community (KBIC), this research tackles a literature gap on applying Traditional Ecological Knowledge (TEK) to forest restoration, an experiential gap in understanding differential climate impacts, and a governance gap perpetuated by colonial metrics that devalue relational ecosystem health.
A multi-method approach systematically addresses these gaps. First, a literature review confirms TEK\u27s application in restoration is under-researched and often extractive. Second, a comparative survey of 232 Tribal and non-Tribal community members provides empirical evidence that climate change disproportionately harms culturally vital Indigenous forest practices, such as gathering medicines and ceremonial materials. Third, a qualitative case study of the KBIC fish hatchery illustrates a living model of relational stewardship guided by reciprocity and seven-generation thinking. Finally, this work proposes the Nature Quotient (NQ) Index, a novel framework to decolonize environmental metrics by valuing cultural connections alongside ecological health.
Integrating these findings, this dissertation demonstrates that climate vulnerability is mediated by cultural relationships and argues for a paradigm shift from managing resources to nurturing relationships. The research provides theoretical, empirical, and practical insights for an environmental governance that honors Indigenous sovereignty, integrates diverse knowledge systems, and fosters just, sustainable human-nature relationships
MACHINE LEARNING DESIGN AND EXPERIMENTAL PRODUCTION OF RECYCLED ALUMINUM ALLOYS WITH MAGNESIUM AND IRON
No full textRecycling aluminum saves up to 95% of the energy required for primary production but introduces impurities such as iron that limit alloy performance. The Al-Mg-Fe system, though commercially used, remains underexplored. This work applies machine learning with Multi-Objective Bayesian Optimization (MOBO) using Thermo-Calc simulations to design Al-Mg-Fe-Si-Cu-Zn alloys with improved castability and reduced porosity. The optimization targeted freezing range, shrinkage, drag coefficient, and microstructural parameters, identifying compositions with high Mg, Si, and Cu and low Zn as optimal. Predicted alloys were experimentally cast and characterized, with yield strengths exceeding 95 MPa and elongation comparable to commercial 319 aluminum in the as- cast state
Batteryless NFC-Enabled Wireless Sensor Node: Design, Optimization, and Implementation
No full textThis thesis presents the design, construction, and testing of a batteryless Near-field communication (NFC) powered wireless sensor node intended for maintenance free, short range Internet of Things (IoT) applications. The work focuses on harvesting energy from a 13.56 MHz NFC field to power a very low-power sensing platform capa- ble of measuring temperature, pressure, and humidity without the use of batteries or wired power. The motivation behind this approach is the growing need for reliable, sustainable, and low-maintenance sensing systems that can operate in environments where battery replacement is impractical, undesirable, or environmentally costly. The prototype is built around a single integrated NFC chip that combines the NFC interface, on-chip memory, and an embedded low power processor. This sin- gle chip solution integrates the NFC front end, a tuned loop antenna for energy harvesting, multiple sensor connection options, and a control program designed for operation from intermittently available harvested energy. The completed prototype performs on demand data acquisition, stores measure- ments in on chip non volatile memory, and communicates results through a standards compliant NFC link while powered entirely by the reader’s Radio Frequency(RF) field. Key engineering contributions include practical antenna tuning for improved harvested voltage, power-aware sensor integration strategies, and development of a simple control flow that keeps active time short enough to remain within the available energy budget. Experimental evaluation confirms that the implemented NFC powered design can reliably acquire sensor data and communicate with an NFC reader within practical near-field distances. The results demonstrate achievable trade offs between energy availability, sensing delay, memory usage, and communication stability, and show that batteryless NFC sensing is a realistic option for compact, low-maintenance IoT devices
NUMERICAL MODELLING AND CO-OPTIMIZATION OF GASOLINE FUELS FOR GASOLINE COMPRESSION IGNITION USING MULTI COMPONENT APPROACH
No full textFuture mobility is expected to rely on a broad spectrum of powertrain technologies, including battery electric vehicles (BEVs), hybrid electric vehicles (HEVs), fuel cell electric vehicles (FCEVs) and traditional internal combustion engine (ICE) vehicles. Despite the shift toward electrification, internal combustion engines are projected to remain a key component of future powertrains, either as the primary power source or as range extenders to generate electricity in hybrid systems. As such, significant research efforts continue to focus on enhancing the efficiency and reducing the emissions of ICEs to meet increasingly stringent regulatory and environmental targets. Gasoline compression ignition (GCI) has been identified as a promising combustion strategy that can combine the high thermal efficiency of diesel engines with the potential for much lower nitrogen oxides (NOx) and particulate matter (PM) emissions. GCI takes advantage of the lower reactivity of gasoline to enable controlled autoignition at lower temperatures and pressures, which in turn reduces the formation of harmful pollutants. However, successful implementation of GCI technology depends not only on engine hardware but also on the chemical and physical properties of the fuel.
This research focuses on understanding the effects of gasoline fuel properties, specifically fuel reactivity on GCI combustion. A range of gasoline fuel formulations were considered, with research octane numbers (RON) from 60 to 90, in order to capture a broad spectrum of autoignition characteristics. In addition to these fuels, three oxygenated blends of ethanol (E36Gas) and isobutanol (iB25gas, iB51Gas) were also studied, along with a baseline RON 87 E10 gasoline, to assess the impact of fuel oxygenation on combustion and emissions. Numerical simulations were conducted using MTU-MRNT, an in-house multidimensional computational fluid dynamics (CFD) solver. The code was coupled with advanced physical sub-models and the Chemkin library to enable accurate modeling of combustion and emissions. Reduced chemical kinetics mechanisms, specifically developed for multicomponent gasoline surrogate fuels, were employed to simulate the oxidation behavior of both conventional and oxygenated fuels.
Initial validation of the fuel models was carried out in a constant volume combustion chamber (CVCC) configuration. The ignition delay and heat release trends observed in simulations closely matched those from experimental data, thereby confirming the model’s capability to predict ignition behavior. Subsequently, a series of parametric studies were performed to explore the sensitivity of ignition to variations in ambient gas density, ambient temperature, fuel injection pressure and oxygen concentration.
Following validation, the fuel models were applied to simulate low-load GCI engine operation at 5 bar brake mean effective pressure (BMEP). The simulation results showed that ignition characteristics, including ignition delay and combustion phasing, exhibited a strong correlation with the fuel\u27s RON. Fuels with lower RONs (higher reactivity) ignited earlier, while higher-RON fuels demonstrated delayed autoignition. This trend aligns well with the experimental measurements
Effects of microbially induced calcite precipitation (MICP) on the soil-concrete interface behavior
No full textImproving the shear strength at the soil-concrete interface can enhance the capacity of structures in various applications, such as horizontal sliding of footings, retaining walls, side friction capacity of pile foundations, and seismic design of tunnel linings. Choosing a sustainable, economical, and environmentally friendly approach is necessary when it comes to engineering practices. To achieve such an approach, the effectiveness of microbially induced calcite precipitation (MICP) in improving the soil-concrete interfacial shear strength was investigated in this study. Soil samples in contact with concrete were treated using the two major MICP pathways, i.e., ureolysis and denitrification. The required microorganisms for both pathways were cultivated from the activated sludge obtained from a wastewater treatment facility. MICP treatment can potentially improve the interface properties through different mechanisms, i.e., bonding the soil particles to the concrete, changing the roughness of the soil particles, and filling the pores between the soil and concrete. The effectiveness of the treatment would also depend on the soil gradation. Therefore, a poorly graded dune sand and a well-graded silty sand were used in this study. A series of laboratory direct shear tests were conducted on both treated and untreated samples. The results of the tests are presented and the effects of MICP pathways, soil gradation, and different contributing mechanisms are discussed
Machine Learning Prediction of Urban Heat Island Severity in the Midwestern United States
Get PDFRapid population growth and urbanization have greatly impacted the environment, causing a sharp rise in city temperatures—a phenomenon known as the Urban Heat Island (UHI) effect. While previous research has extensively examined the influence of land use characteristics on urban heat islands, their impact on community demographics and UHI severity remains unexplored. Moreover, most previous studies have focused on specific locations, resulting in relatively homogeneous environmental data and limiting understanding of variations across different areas. To address this gap, this paper develops ensemble learning models to predict UHI severity based on demographic, meteorological, and land use/land cover factors in Midwestern United States. Analyzing over 11,000 data points from urban census tracts across more than 12 states in the Midwestern United States, this study developed Random Forest and XGBoost classifiers achieving weighted F1-scores up to 0.76 and excellent discriminatory power (ROC-AUC \u3e 0.90). Feature importance analysis, supported by a detailed SHAP (SHapley Additive exPlanations) interpretation, revealed that the difference in vegetation between urban and rural areas (DelNDVI_summer) and imperviousness were the most critical predictors of UHI severity. This work provides a robust, large-scale predictive tool that helps urban planners and policymakers identify key UHI drivers and develop targeted mitigation strategies
Hourly Simulated Power Production Data with Snow Loss Model at Queued Utility-Scale PV Sites Simulated as Fixed-Tilt Systems in the U.S. Eastern Interconnection for Weather Year 2016
No full textUsing 2016 weather data, we ran PySAM power production simulations for utility-scale PV sites in the U.S. Eastern Interconnection queue. Site IDs, capacities, and locations (counties) were extracted from Lawrence Berkeley National Laboratory’s Queued Up: 2024 Edition dataset. No panel mount information was provided, so all sites were assumed to be 30-degree, fixed tilt systems. Sites’ latitudes and longitudes were assumed to be the centers of the installation counties. See queued_site_metadata.csv file for individual site metadata