26800 research outputs found
Sort by
Time-invariant characteristics of evaporating thin film during droplet evaporation
No full textAn evaporating thin film (ETF) plays a critical role in heat dissipation due to the substantially enhanced evaporation flux. Despite its fundamental importance in interfacial evaporation dynamics, direct experimental characterization of the ETF has remained a major challenge, resulting in a limited understanding of its temporal evolution. This study provides the first direct experimental evidence demonstrating the time-invariant nature of ETF profiles during the pinning stage of droplet evaporation, offering novel insight into interfacial evaporation mechanisms. The ultra-thin liquid film profiles are measured under varying surface wettability conditions through a self-assembled monolayer (SAM) on a gold substrate. A theoretical model incorporating kinetic theory and the augmented Young-Laplace equation is also employed to quantitatively examine the finite evaporation flux in the ETF region. The findings reveal that the ETF profile remains time-invariant during the pinned stage, exhibiting local equilibrium characteristics, and theoretical predictions show good agreement with experimental observations. Notably, it is found that surface wettability has a minor influence on variations in ETF profiles, and the disjoining pressure is found to be higher on hydrophilic substrates, contributing to further thinning of the ETF
Defect Correction Methods for Fluid Flows at High Reynolds Numbers
No full textDefect Correction Methods for Fluid Flows at High Reynold\u27s Numbers presents the mathematical development of defect correction methods (DCM) in application to fluid flow problems in various settings. We will show several approaches to applying the DCM ideas in computational fluid dynamics (CFD) – from a basic idea of controlling the flow by the means of increased diffusion, to the state-of-the-art family of novel, DCM-based turbulence models. The main idea of the methods presented in this book, is to use defect correction in turbulence modelling; additionally, several methods will also be presented, that aim at reducing the time discretization error. Features · Provides a road map, starting from the ideas of minimally invasive controlling of turbulent flows, to the ways of improving the existing regularization techniques with DCM, to the ideas of \u27full defect correction\u27 in both space and time and, finally, to the more complex embedding of the DCM into turbulence modelling by the \u27correction\u27 of the whole turbulence model · Can be used for teaching a topics course on a Masters or Ph.D. level. It is even more suitable as a reference for CFD theorists and practitioners, with most of the methods being minimally invasive and, therefore, easy to implement in the existing/legacy codes · Discusses the current challenges in turbulence modelling with defect correction, showing several possible directions for future developments. Two source codes are provided – one for a regularization technique and another for a novel turbulence model – in order to give an interested researcher a quick start to the topic of DCM in CFD
Replicating associative learning of rodents with a neuromorphic robot in an open-field arena
No full textThis study emulates associative learning in rodents by using a neuromorphic robot navigating an open-field arena. The goal is to investigate how biologically inspired neural models can reproduce animal-like learning behaviors in real-world robotic systems. We constructed a neuromorphic robot by deploying computational models of spatial and sensory neurons onto a mobile platform. Different coding schemes—rate coding for vibration signals and population coding for visual signals—were implemented. The associative learning model employs 19 spiking neurons and follows Hebbian plasticity principles to associate visual cues with favorable or unfavorable locations. Our robot successfully replicated classical rodent associative learning behavior by memorizing causal relationships between environmental cues and spatial outcomes. The robot’s self-learning capability emerged from repeated exposure and synaptic weight adaptation, without the need for labeled training data. Experiments confirmed functional learning behavior across multiple trials. This work provides a novel embodied platform for memory and learning research beyond traditional animal models. By embedding biologically inspired learning mechanisms into a real robot, we demonstrate how spatial memory can be formed and expressed through sensorimotor interactions. The model’s compact structure (19 neurons) illustrates a minimal yet functional learning network, and the study outlines principles for synaptic weight and threshold design, guiding future development of more complex neuromorphic systems
Hourly Simulated Power Production Data with Snow Loss Model at Existing Utility-Scale PV Sites (\u3e5 MW) in the U.S. Eastern Interconnection in 2019
No full textProject Summary: We ran PySAM power production simulations for utility-scale (\u3e5 MW) PV sites located in the U.S. Eastern Interconnection in the year 2019. Site panel mounts (fixed-tilt or single-axis tracking), capacities, and locations (latitudes and longitudes) were extracted from Lawrence Berkeley National Laboratory\u27s Utility-Scale Solar 2024 Edition dataset. See 2019_PV_existing_site_metadata.csv file for individual site metadata
Hourly Simulated Power Production Data with No Snow Loss Model at Existing Utility-Scale PV Sites (\u3e5 MW) in the U.S. Eastern Interconnection in 2017
No full textProject Summary: We ran PySAM power production simulations for utility-scale (\u3e5 MW) PV sites located in the U.S. Eastern Interconnection in the year 2017. Site panel mounts (fixed-tilt or single-axis tracking), capacities, and locations (latitudes and longitudes) were extracted from Lawrence Berkeley National Laboratory\u27s Utility-Scale Solar 2024 Edition dataset. See 2017_PV_existing_site_metadata.csv file for individual site metadata
Comparison of In Situ Plant Area Index and Remotely Sensed Leaf Area Index of Northeastern American Deciduous, Mixed, and Coniferous Forests for SMAPVEX19-22
No full textLeaf Area Index (LAI) and Plant Area Index (PAI) are critical biophysical parameters for quantifying foliage density of crops and trees. As such they may also be effective metrics to estimate the impact of leafy vegetation on microwave signals, like the radiometer on board NASA’s Soil Moisture Active Passive (SMAP) satellite. This study presents a comprehensive analysis of PAI and LAI measurements acquired through in situ and remotely sensed (RS) methods in diverse forest ecosystems of northeastern America, including deciduous, mixed, and coniferous forests. We compare RS LAI and ground-truth PAI data to understand the variation between the RS LAI products and the impacts of RS spatial resolution on analysis. We perform these comparisons in the context of understanding SMAP’s Vegetation Optical Depth (VOD) retrievals, and compare RS LAI with SMAP VOD measurements. We find strong (R2 \u3e 0.83) positive relationships between in situ PAI and the tested RS LAI products, with improved positive relationships (R2 \u3e 0.92) for higher spatial resolution (\u3c 30m) RS LAI products. We also discuss considerations for the LAI algorithms. The higher resolution RS products showed a consistently low bias (by ~1 unit) in both spring and summer compared to in situ PAI measurements, while the coarse resolution LAI products matched in situ PAI values in spring but overestimated LAI by ~ 1 unit in the summer. Smaller y-intercept values (\u3c -0.56) associated with the higher resolution products relative to the coarse resolution products (\u3e -0.29) could indicate greater influence of woody biomass on higher resolution RS LAI algorithms relative to low resolution RS LAI algorithms. Comparisons of RS LAI and VOD showed generally positive relationships that varied by satellite sensor
A Survey on the Scheduling of DL and LLM Training Jobs in GPU Clusters
No full textAs deep learning (DL) technology rapidly advances in areas such as computer vision, natural language processing, and more recently, large language models (LLMs), the demand for computing resources has increasingly grown. In particular, scheduling deep learning training (DLT) jobs on graphics processing unit (GPU) clusters has become crucial for the effective utilization of computing resources and the acceleration of model training processes. However, resource management and scheduling in GPU clusters face challenges related to computing and communication, including job sharing, interference, elastic scheduling, heterogeneous resources, and fairness. This survey investigates the scheduling issues of DLT jobs in GPU clusters, focusing on scheduling optimizations at the job characteristic and cluster resource levels. We analyze the structure and training computing characteristics of traditional DL models and LLMs, as well as their requirements for iterative computation, communication, GPU sharing, and resource elasticity. In addition, we compare the main contributions of this survey with related reviews and discuss research directions, including scheduling based on job characteristics and optimization strategies for cluster resources. This survey aims to provide researchers and practitioners with a comprehensive understanding of DLT job scheduling in GPU clusters and to point out directions for future research
Potential for contaminant biotransport by migratory fish prior to dam removal and selective fish passage in a Great Lakes tributary
No full textDam removals and fish passage can enhance aquatic connectivity but may also promote upstream transport of legacy contaminants by migratory fish. This study assessed the potential for contaminant biotransport in Michigan\u27s Boardman River following the planned removal of the Union Street Dam and installation of FishPass, a selective fish passage facility. We quantified polychlorinated biphenyls (PCBs), mercury (Hg), and organochlorine pesticides in carcass and egg samples from migratory species including Chinook and coho salmon, migratory rainbow trout, common white and longnose sucker, lake trout, walleye, and sea lamprey. Chinook salmon exhibited the highest PCB concentrations in both carcasses and eggs, exceeding those of rainbow trout and native suckers. Similarly, Chinook salmon were predicted to deposit up to 2,200 mg of PCBs upstream under a high run size scenario-over 80 and 100 times greater than rainbow trout and native suckers. While suckers had lower individual contaminant burdens, their relatively large run sizes contributed moderately to potential contaminant biotransport compared to rainbow trout indicating an interaction between abundance and spawner contaminant burden. Stream-resident brook and brown trout in reaches open to migratory fish had higher PCBs and lower Hg concentrations than in closed reaches, likely reflecting dietary exposure to eggs and growth dilution. These results demonstrate that the potential for contaminant biotransport varies widely among migratory species and highlights the need for managers to consider both contaminant burden and run size when making fish passage decisions to balance ecological restoration with contaminant exposure risk
A lumped-parameter model for the mechanics of interlocked geometries: Ribbons, origami, and woven structures
No full textThis work develops a lumped-parameter mechanical simulation to investigate the mechanics of interlocking behaviors in thin membrane systems like origami, ribbons, and woven structures. Unlike traditional finite element models (FEM) that require fine meshes and long computation to obtain good results, lumped-parameter models use specially derived coarse mesh formulations for rapid computation. Traditional lumped-parameter models tend to use a truss-based formulation to represent membrane systems, which can produce less accurate computation of in-plane stiffness. Moreover, these models lack the ability to capture contact behaviors in interlocked geometries. To address the challenge, this work develops a new lumped-parameter simulation with three components: a four-node rotational spring element to capture out-of-plane bending and folding, a new large-deformation triangle element to capture in-plane stretching and shearing, and a new triangle-to-triangle contact element for interlocking behaviors. Using verification tests, this work shows that the proposed model can outperform other lumped-parameter models in predicting the mechanics of thin membranes with potential contact. In addition, practical examples are presented to demonstrate the effectiveness of the simulation framework. Our results show that the proposed model can be 25 times faster than a full FEM simulation implemented in Abaqus for a selected case study. Finally, an implementation software package is developed to execute the proposed simulation and is published with this article
INTELLIGENT SYSTEMS FOR SYNERGISTIC OPTIMIZATION OF VEHICLE DYNAMICS AND ENERGY CONSUMPTION IN CONNECTED AND AUTOMATED VEHICLES
No full textThis dissertation focuses on the development and validation of on-board, real-time estimation algorithms and the investigation of energy-saving strategies for modern light-duty vehicles, including connected and automated vehicles (CAVs). The accurate estimation of key vehicle parameters is critical for enhancing energy efficiency, enabling advanced driver-assistance systems (ADAS), and optimizing vehicle performance and energy efficiency. First, this research introduces a novel real-time algorithm to dynamically learn vehicle mass using readily available sensor data. Based on longitudinal vehicle dynamics, a sensitivity analysis was conducted to identify the conditions under which mass estimation is most robust. The algorithm was extensively validated through on-road testing with multiple electric and plug-in hybrid vehicles, demonstrating high accuracy and repeatability against weighed vehicle mass across various loading conditions. Second, an algorithm for real-time road load estimation was developed to determine the forces resisting a vehicle\u27s motion, such as aerodynamic drag and rolling resistance. By deriving analytical solutions from the equations of motion, the algorithm effectively learns road load coefficients from normal driving data. The methodology was validated using simulations, controlled coast-down experiments on a dynamometer and on-road, and real-world driving tests, showing a significant improvement in the accuracy of energy consumption predictions. Building on these estimation capabilities, the dissertation investigates the energy-saving potential of automated vehicle-following. Through extensive on-road experiments with two distinct vehicles, the study quantifies the reduction in energy consumption for a following vehicle due to aerodynamic drafting. The research systematically analyzes the effects of speed, longitudinal gap, and lateral offset on the energy savings for both the lead and following vehicles. Finally, a theoretical framework for optimizing a vehicle\u27s speed trajectory at signalized intersections is presented. Using a variational calculus approach, the study formulates and solves the problem of minimizing axle energy for Eco-Approach and Departure maneuvers, laying the groundwork for advanced, energy-efficient control systems in CAVs. Collectively, this work provides a suite of validated intelligent systems and methodologies that can significantly improve the energy efficiency and performance of next-generation automobiles