Hasil untuk "physics.flu-dyn"

Yangning Li, Yinghui Li, Xinyu Wang et al.

Multimodal Retrieval Augmented Generation (mRAG) plays an important role in mitigating the"hallucination"issue inherent in multimodal large language models (MLLMs). Although promising, existing heuristic mRAGs typically predefined fixed retrieval processes, which causes two issues: (1) Non-adaptive Retrieval Queries. (2) Overloaded Retrieval Queries. However, these flaws cannot be adequately reflected by current knowledge-seeking visual question answering (VQA) datasets, since the most required knowledge can be readily obtained with a standard two-step retrieval. To bridge the dataset gap, we first construct Dyn-VQA dataset, consisting of three types of"dynamic"questions, which require complex knowledge retrieval strategies variable in query, tool, and time: (1) Questions with rapidly changing answers. (2) Questions requiring multi-modal knowledge. (3) Multi-hop questions. Experiments on Dyn-VQA reveal that existing heuristic mRAGs struggle to provide sufficient and precisely relevant knowledge for dynamic questions due to their rigid retrieval processes. Hence, we further propose the first self-adaptive planning agent for multimodal retrieval, OmniSearch. The underlying idea is to emulate the human behavior in question solution which dynamically decomposes complex multimodal questions into sub-question chains with retrieval action. Extensive experiments prove the effectiveness of our OmniSearch, also provide direction for advancing mRAG. The code and dataset will be open-sourced at https://github.com/Alibaba-NLP/OmniSearch.

Koji Fukagata, Kai Fukami

An autoencoder is a self-supervised machine-learning network trained to output a quantity identical to the input. Owing to its structure possessing a bottleneck with a lower dimension, an autoencoder works to achieve data compression, extracting the essence of the high-dimensional data into the resulting latent space. We review the fundamentals of flow field compression using convolutional neural network-based autoencoder (CNN-AE) and its applications to various fluid dynamics problems. We cover the structure and the working principle of CNN-AE with an example of unsteady flows while examining the theoretical similarities between linear and nonlinear compression techniques. Representative applications of CNN-AE to various flow problems, such as mode decomposition, latent modeling, and flow control, are discussed. Throughout the present review, we show how the outcomes from the nonlinear machine-learning-based compression may support modeling and understanding a range of fluid mechanics problems.

M. Ehsan, M. L’Heureux, M. Tippett et al.

This paper provides an updated assessment of the “International Research Institute for Climate and Society’s (IRI) El Niño Southern Oscillation (ENSO) Predictions Plume”. We evaluate 253 real-time forecasts of the Niño 3.4 index issued from February 2002 to February 2023 and examine multimodal means of dynamical (DYN) and statistical (STAT) models separately. Forecast skill diminishes as lead time increases in both DYN and STAT forecasts, with peak accuracy occurring post-northern hemisphere spring predictability barrier and preceding seasons. The DYN forecasts outperform STAT forecasts with a pronounced advantage in forecasts initiated from late boreal winter through spring. The analysis uncovers an asymmetry in predicting the onset of cold and warm ENSO episodes, with warm episode onsets being better forecasted than cold onsets in both DYN and STAT models. The DYN forecasts are found to be valuable for predicting warm and cold ENSO episode onsets at least several months in advance, while STAT forecasts are less informative about ENSO phase transitions. The results indicate that predicting ENSO onset is challenging and that the ability to do so is both model- and event-dependent.

Zhihang Yu, Yiqun Chen, Jingjing Li et al.

Pathological conditions linked to shear stress have been identified in hematological diseases, cardiovascular diseases, and cancer. These conditions often exhibit significantly elevated shear stress levels, surpassing 1000 dyn/cm2 in severely stenotic arteries. Heightened shear stress can induce mechanical harm to endothelial cells, potentially leading to bleeding and fatal consequences. However, current technology still grapples with limitations, including inadequate flexibility in simulating bodily shear stress environments, limited range of shear stress generation, and spatial and temporal adaptability. Consequently, a comprehensive understanding of the mechanisms underlying the impact of shear stress on physiological and pathological conditions, like thrombosis, remains inadequate. To address these limitations, this study presents a microfluidic-based shear stress generation chip as a proposed solution. The chip achieves a substantial 929-fold variation in shear stress solely by adjusting the degree of constriction in branch channels after PDMS fabrication. Experiments demonstrated that a rapid increase in shear stress up to 1000 dyn/cm2 significantly detached 88.2% cells from the substrate. Long-term exposure (24 h) to shear stress levels below 8.3 dyn/cm2 did not significantly impact cell growth. Furthermore, cells exposed to shear stress levels equal to or greater than 8.3 dyn/cm2 exhibited significant alterations in aspect ratio and orientation, following a normal distribution. This microfluidic chip provides a reliable tool for investigating cellular responses to the wide-ranging shear stress existing in both physiological and pathological flow conditions.

Huikun Wang, Rodolfo J. Flores, Hector E. Yarur et al.

Prefrontal cortical (PFC) circuits provide top-down control of threat reactivity. This includes ventromedial PFC (vmPFC) circuitry, which plays a role in suppressing fear-related behavioral states. Dynorphin (Dyn) has been implicated in mediating negative affect and maladaptive behaviors induced by severe threats and is expressed in limbic circuits, including the vmPFC. However, there is a critical knowledge gap in our understanding of how vmPFC Dyn-expressing neurons and Dyn transmission detect threats and regulate expression of defensive behaviors. Here, we demonstrate that Dyn cells are broadly activated by threats and release Dyn locally in the vmPFC to limit passive defensive behaviors. We further demonstrate that vmPFC Dyn-mediated signaling promotes a switch of vmPFC networks to a fear-related state. In conclusion, we reveal a previously unknown role of vmPFC Dyn neurons and Dyn neuropeptidergic transmission in suppressing defensive behaviors in response to threats via state-driven changes in vmPFC networks.

Aida Mohammadkhani, Min Qiao, S. Borgland

Dopamine (DA) neurons in the ventral tegmental area (VTA) respond to motivationally relevant cues, and circuit-specific signaling drives different aspects of motivated behavior. Orexin (ox; also known as hypocretin) and dynorphin (dyn) are coexpressed lateral hypothalamic (LH) neuropeptides that project to the VTA. These peptides have opposing effects on the firing activity of VTADA neurons via orexin 1 (Ox1R) or kappa opioid (KOR) receptors. Given that Ox1R activation increases VTADA firing, and KOR decreases firing, it is unclear how the coreleased peptides contribute to the net activity of DA neurons. We tested if optical stimulation of LHox/dyn neuromodulates VTADA neuronal activity via peptide release and if the effects of optically driven LHox/dyn release segregate based on VTADA projection targets including the basolateral amygdala (BLA) or the lateral or medial shell of the nucleus accumbens (lAcbSh, mAchSh). Using a combination of circuit tracing, optogenetics, and patch-clamp electrophysiology in male and female orexincre mice, we showed a diverse response of LHox/dyn optical stimulation on VTADA neuronal firing, which is not mediated by fast transmitter release and is blocked by antagonists to KOR and Ox1R signaling. Additionally, where optical stimulation of LHox/dyn inputs in the VTA inhibited firing of the majority of BLA-projecting VTADA neurons, optical stimulation of LHox/dyn inputs in the VTA bidirectionally affects firing of either lAcbSh- or mAchSh-projecting VTADA neurons. These findings indicate that LHox/dyn corelease may influence the output of the VTA by balancing ensembles of neurons within each population which contribute to different aspects of reward seeking.

Yizeng Han, Dongchen Han, Zeyu Liu et al.

Early exiting has become a promising approach to improving the inference efficiency of deep networks. By structuring models with multiple classifiers (exits), predictions for "easy" samples can be generated at earlier exits, negating the need for executing deeper layers. Current multi-exit networks typically implement linear classifiers at intermediate layers, compelling low-level features to encapsulate high-level semantics. This sub-optimal design invariably undermines the performance of later exits. In this paper, we propose Dynamic Perceiver (Dyn-Perceiver) to decouple the feature extraction procedure and the early classification task with a novel dual-branch architecture. A feature branch serves to extract image features, while a classification branch processes a latent code assigned for classification tasks. Bi-directional cross-attention layers are established to progressively fuse the information of both branches. Early exits are placed exclusively within the classification branch, thus eliminating the need for linear separability in low-level features. Dyn-Perceiver constitutes a versatile and adaptable framework that can be built upon various architectures. Experiments on image classification, action recognition, and object detection demonstrate that our method significantly improves the inference efficiency of different backbones, outperforming numerous competitive approaches across a broad range of computational budgets. Evaluation on both CPU and GPU platforms substantiate the superior practical efficiency of Dyn-Perceiver. Code is available at https://www.github.com/LeapLabTHU/Dynamic_Perceiver.

Richard Jeske, Chang Liu, L. Duke et al.

Human mesenchymal stromal cells (hMSCs) are mechanically sensitive undergoing phenotypic alterations when subjected to shear stress, cell aggregation, and substrate changes encountered in 3D dynamic bioreactor cultures. However, little is known about how bioreactor microenvironment affects the secretion and cargo profiles of hMSC-derived extracellular vesicles (EVs) including the subset, "exosomes", which contain therapeutic proteins, nucleic acids, and lipids from the parent cells. In this study, bone marrow-derived hMSCs were expanded on 3D Synthemax II microcarriers in the PBS mini 0.1L Vertical-Wheel bioreactor system under variable shear stress levels at 25, 40, and 64 RPM (0.1-0.3 dyn/cm2). The bioreactor system promotes EV secretion from hMSCs by 2.5-fold and upregulates the expression of EV biogenesis markers and glycolysis genes compared to the static 2D culture. The microRNA cargo was also altered in the EVs from bioreactor culture including the upregulation of miR-10, 19a, 19b, 21, 132, and 377. EV protein cargo was characterized by proteomics analysis, showing upregulation of metabolic, autophagy and ROS-related proteins comparing with 2D cultured EVs. In addition, the scalability of the Vertical-Wheel bioreactor system was demonstrated in a 0.5L bioreactor, showing similar or better hMSC-EV secretion and cargo content compared to the 0.1L bioreactor. This study advances our understanding of bio-manufacturing of stem cell-derived EVs for applications in cell-free therapy towards treating neurological disorders such as ischemic stroke, Alzheimer's disease, and multiple sclerosis.

Bixuan Wang, Ruizhe Zhang, Hengyu Wang et al.

We employ a new circuit method to characterize the gate dynamic breakdown voltage ( $\textit {BV}_{\text {dyn}}$ ) of Schottky-type p-gate GaN HEMTs in power converters. Different from prior pulse I-V and DC stress tests, this method features a resonance-like gate ringing with the pulse width down to 20 ns and an inductive switching concurrently in the drain-source loop. At the increased pulse width, the gate $\textit {BV}_{\text {dyn}}$ shows a decrease and then saturation at 21~22 V. Moreover, the gate $\textit {BV}_{\text {dyn}}$ increases with temperature and is higher under the hard switching than that under the drain-source grounding condition. In the 400 V hard switching at 150 °C, the gate $\textit {BV}_{\text {dyn}}$ reaches 27.5 V. Such impact of the drain switching locus and temperature on the gate $\textit {BV}_{\text {dyn}}$ is not seen in Si and SiC power transistors tested in the same setup. These results are explained by a physics model that accounts for the electrostatics in the p-GaN gate stack in hard switching and at high temperatures. This work unveils new physics critical to the gate robustness of p-gate GaN HEMTs and manifest the necessity of the gate robustness evaluation in inductive switching conditions.

J. Vidal, D. Cébron

Motivated by modelling rotating turbulence in planetary fluid layers, we investigate precession-driven flows in ellipsoids subject to stress-free boundary conditions (SF-BC). The SF-BC could indeed unlock numerical constraints associated with the no-slip boundary conditions (NS-BC), but are also relevant for some astrophysical applications. Although SF-BC have been employed in the pioneering work of Lorenzani & Tilgner (J. Fluid Mech., 2003, 492, pp. 363--379), they have scarcely been used due to the discovery of some specific mathematical issues associated with angular momentum conservation. We revisit the problem using asymptotic analysis in the low-viscosity regime, which is validated with numerical simulations. First, we extend the reduced model of uniform-vorticity flows in ellipsoids to account for SF-BC. We show that the long-term evolution of angular momentum is affected by viscosity in triaxial geometries, but also in axisymmetric ellipsoids when the mean rotation axis of the fluid is not the symmetry axis. In a regime relevant to planets, we analytically obtain the primary forced flow in triaxial geometries, which exhibits a second inviscid resonance. Then, we investigate the bulk instabilities existing in precessing ellipsoids. We show that using SF-BC would be useful to explore the non-viscous instabilities (e.g. Kerswell, Geophys. Astrophys. Fluid Dyn., 1993, 72, pp. 107-144), which are presumably relevant for planetary applications but are often hampered in experiments or simulations with NS-BC.

Eshan Sharma

Numerical simulations of a partially-premixed, turbulent jet diffusion flame stabilised in a hot vitiated co-flow are performed. For auto-igniting flames, an accurate prediction of flame stabilisation, which depends on a delicate balance between turbulent transport and chemical kinetics at the flame base, poses an enormous challenge to conventional turbulent combustion models. Multiple mapping conditioning/large eddy simulation (MMC-LES), a promising tool for modelling turbulence-chemistry interactions, has been successfully applied to simulate a variety of combustion applications involving gaseous, liquid, and solid fuels. MMC-LES is a full \gls{PDF} method where MMC plays the role of the mixing model, emulating molecular mixing phenomenon. MMC attempts to produce accurate molecular mixing by localising mixing in an independent, composition-like reference space. Due to this enforced localised mixing, a sparse distribution of stochastic Lagrangian particles for may be used for the Monte-Carlo simulation of the sub-grid joint-composition PDF equation. In the present study, we employ MMC-LES to investigate the auto-igniting methane/air flames of UC Berkeley. A sparse resolution of 1 particle per 10 Eulerian finite-volume cells is used in this study, which offers much cheaper computing expenses in comparison to the conventional transported PDF approach. A skeletal chemical mechanism, based on GRI 3.0, containing 30 species and 184 reactions, represents the oxidation of methane. The time-scale of molecular mixing is modelled using the recently published \textit{dyn-aISO} model to assess its performance in an auto-igniting configuration.

Guoquan Qina, Zhenya Yan, Boling Guo

This paper investigates the Cauchy problem of a generalized Camassa-Holm equation with quadratic and cubic nonlinearities (alias the mCH-Novikov-CH equation), which is a generalization of some special equations such as the Camassa-Holm (CH) equation, the modified CH (mCH) equation ((alias the Fokas-Olver-Rosenau-Qiao equation), the Novikov equation, the CH-mCH equation, the mCH-Novikov equation, and the CH-Novikov equation. We first show the local well-posedness for the strong solutions of the mCH-Novikov-CH equation in Besov spaces by means of the Littlewood-Paley theory and the transport equations theory. Then, the Holder continuity of the data-to-solution map to this equation are exhibited in some Sobolev spaces. After providing the blow-up criterion and the precise blow-up quantity in light of the Moser-type estimate in the Sobolev spaces, we then trace a portion and the whole of the precise blow-up quantity, respectively, along the characteristics associated with this equation, and obtain two kinds of sufficient conditions on the gradient of the initial data to guarantee the occurance of the wave-breaking phenomenon. Finally, the non-periodic and periodic peakon and multi-peakon solutions for this equation are also explored.

Hugh J. A. Bird, Kiran Ramesh

Frequency domain Unsteady Lifting-Line Theory (ULLT) provides a means by which the aerodynamics of oscillating wings may be studied at low computational cost without neglecting the interacting effects of aspect ratio and oscillation frequency. Renewed interest in the method has drawn attention to several uncertainties however. Firstly, to what extent is ULLT practically useful for rectangular wings, despite theoretical limitations? And secondly, to what extent is a complicated wake model needed in the outer solution for good accuracy? This paper aims to answer these questions by presenting a complete ULLT based on the work of Sclavounos, along with a novel ULLT that considers only the streamwise vorticity and a Prandtl-like pseudosteady ULLT. These are compared to high fidelity Euler CFD for cases of rectangular wings at multiple aspect ratios and oscillation frequencies. The results of this work establish ULLT as a low computational cost model capable of accounting for interacting finite-wing and oscillation frequency effects, and identify the aspect ratio and frequency regimes where the three ULLTs are most accurate. This research paves the way towards the construction of time-domain or numerical ULLTs which may be augmented to account for non-linearities such as flow separation.

Sebastien Galtier

Capillary waves are perhaps the simplest example to consider for an introduction to wave turbulence. Since the first paper by Zakharov and Filonenko (1967), capillary wave turbulence has been the subject of many studies but a didactic derivation of the kinetic equation is still lacking. It is the objective of this paper to present such a derivation in absence of gravity and in the approximation of deep water. We use the Eulerian method and a Taylor expansion around the equilibrium elevation for the velocity potential to derive the kinetic equation. The use of directional polarities for three-wave interactions leads to a compact form for this equation which is fully compatible with previous work. The exact solutions are derived with the so-called Zakharov transformation applied to wavenumbers and the nature of these solutions is discussed. Experimental and numerical works done in recent decades are also reviewed.

Zsolt Farkas, M. Petrič, Xianghua Liu et al.

Phagocytosis of various targets, such as apoptotic cells or opsonized pathogens, by macrophages is coordinated by a complex signaling network initiated by distinct phagocytic receptors. Despite the different initial signaling pathways, each pathway ends up regulating the actin cytoskeletal network, phagosome formation and closure, and phagosome maturation leading to degradation of the engulfed particle. Herein, we describe a new phagocytic function for the nucleoside diphosphate kinase 1 (NDK‐1), the nematode counterpart of the first identified metastasis inhibitor NM23‐H1 (nonmetastatic clone number 23) nonmetastatic clone number 23 or nonmetastatic isoform 1 (NME1). We reveal by coimmunoprecipitation, Duolink proximity ligation assay, and mass spectrometry that NDK‐1/NME1 works in a complex with DYN‐1/Dynamin (Caenorhabditis elegans/human homolog proteins), which is essential for engulf ment and phagosome maturation. Time‐lapse microscopy shows that NDK‐1 is expressed on phagosomal surfaces during cell corpse clearance in the same time window as DYN‐1. Silencing of NM23‐M1 in mouse bone marrow—derived macrophages resulted in decreased phagocytosis of apoptotic thymocytes. In human macrophages, NM23‐H1 and Dynamin are corecruited at sites of phagosome formation in F‐actin—rich cups. In addition, NM23‐H1 was required for efficient phagocytosis. Together, our data demonstrate that NDK‐1/NME1 is an evolutionarily conserved element of successful phagocytosis.—Farkas, Z., Petric, M., Liu, X., Herit, F., Rajnavölgyi, É., Szondy, Z., Budai, Z., Orbán, T. I., Sándor, S., Mehta, A., Bajtay, Z., Kovács, T., Jung, S. Y., Afaq Shakir, M., Qin, J., Zhou, Z., Niedergang, F., Boissan, M., Takács‐Vellai, K. The nucleoside diphosphate kinase NDK‐1/NME1 promotes phagocytosis in concert with DYN‐1/dynamin. FASEB J. 33, 11606–11614 (2019). www.fasebj.org

Leon Rueckert, Annette Zippelius, Reiner Kree

We consider a liquid droplet which is propelled solely by internal flow. In a simple model, this flow is generated by an autonomous actuator, which moves on a prescribed trajectory inside the droplet. In a biological system, the device could represent a motor, carrying cargo and moving on a filamentary track. We work out the general framework to compute the self-propulsion of the droplet as a function of the actuating forces and the trajectory. The simplest autonomous device is composed of three point forces. Such a device gives rise to linear, circular or spiraling motion of the droplet, depending on whether the device is stationary or moving along a radial track. As an example of a more complex track we study in detail a spherical looped helix, inspired by recent studies on the propulsion of Synechococcus1 and Myxobacteria2. The droplet trajectories are found to depend strongly on the orientation of the device and the direction of the forces relative to the track with the posibility of unbounded motion even for time independent forcing.

Florian Zaussinger, Friedrich Kupka

We present a numerical study of double-diffusive convection characterized by a stratification unstable to thermal convection while at the same time a mean molecular weight (or solute concentration) difference between top and bottom counteracts this instability. Convective zones can form in this case either by the stratification being locally unstable to the combined action of both temperature and solute gradients or by another process, the oscillatory double-diffusive convective instability, which is triggered by the faster molecular diffusivity of heat in comparison with that one of the solute. We discuss successive layer formation for this problem in the case of an instantaneously heated bottom (plate) which forms a first layer with an interface that becomes temporarily unstable and triggers the formation of further, secondary layers. We consider both the case of a Prandtl number typical for water (oceanographic scenario) and of a low Prandtl number (giant planet scenario). We discuss the impact of a Couette like shear on the flow and in particular on layer formation for different shear rates. Additional layers form due to the oscillatory double-diffusive convective instability, as is observed for some cases. We also test the physical model underlying our numerical experiments by recovering experimental results of layer formation obtained in laboratory setups.

Andre Giesecke, Tobias Vogt, Thomas Gundrum et al.

The project DRESDYN (DREsden Sodium facility for DYNamo and thermohydraulic studies) conducted at Helmholtz-Zentrum Dresden-Rossendorf (HZDR) provides a new platform for a variety of liquid sodium experiments devoted to problems of geo- and astrophysical magnetohydrodynamics. The most ambitious experiment within this project is a precession driven dynamo experiment that currently is under construction. It consists of a cylinder filled with liquid sodium that simultaneously rotates around two axes. The experiment is motivated by the idea of a precession-driven flow as a complementary energy source for the geodynamo or the ancient lunar dynamo. In the present study we address numerical and experimental examinations in order to identify parameter regions where the onset of magnetic field excitation will be most probable. Both approaches show that in the strongly nonlinear regime the flow is essentially composed of the directly forced primary Kelvin mode and higher modes in terms of standing inertial waves that arise from nonlinear self-interactions. A peculiarity is the resonance-like emergence of an axisymmetric mode that represents a double roll structure in the meridional plane, which, however, only occurs in a very limited range of the precession ratio. This axisymmetric mode turns out to be beneficial for dynamo action, and kinematic simulations of the magnetic field evolution induced by the time-averaged flow exhibit magnetic field excitation at critical magnetic Reynolds numbers around ${\rm{Rm}}^{\rm{c}}\approx 430$, which is well within the range of the planned liquid sodium experiment.

Yali Liu, Shuisheng Li, Qing Wang et al.

M. Bruegel, D. Nagel, M. Funk et al.