arXiv subjects

Yaxing Li

Publications and source records attributed to Yaxing Li.

15 recordsLinked to original sources

arXiv2025

OpenSWI: A Massive-Scale Benchmark Dataset for Surface Wave Dispersion Curve Inversion

Surface wave dispersion curve inversion plays a critical role in both shallow resource exploration and deep geological studies, yet it remains hindered by sensitivity to initial models and low computational efficiency. Recently, data-driven deep learning methods, inspired by advances in computer vision, have shown promising potential to address these challenges. However, the lack of large-scale, diverse benchmark datasets remains a major obstacle to their development and evaluation. To bridge this gap, we present OpenSWI, a comprehensive benchmark dataset generated through the Surface Wave Inversion Dataset Preparation (SWIDP) pipeline. OpenSWI includes two synthetic datasets tailored to different research scales and scenarios, OpenSWI-shallow and OpenSWI-deep, and an AI-ready real-world dataset for generalization evaluation, OpenSWI-real. OpenSWI-shallow, derived from the 2-D OpenFWI geological model dataset, contains over 22 million 1-D velocity profiles paired with fundamental-mode phase and group velocity dispersion curves, spanning a wide range of shallow geological structures (e.g., flat layers, faults, folds, realistic stratigraphy). OpenSWI-deep, built from 14 global and regional 3-D geological models, comprises 1.26 million high-fidelity 1-D velocity-dispersion pairs for deep-Earth studies. OpenSWI-real, compiled from open-source projects, contains two sets of observed dispersion curves with corresponding reference models, serving as a benchmark for evaluating model generalization. To demonstrate utility, we trained models on OpenSWI-shallow and -deep and evaluated them on OpenSWI-real, demonstrating strong agreement between predictions and references, which confirms the diversity and representativeness of the dataset. To advance intelligent surface wave inversion, we release the SWIDP toolbox, OpenSWI datasets, and trained models for the research community.

physics.geo-ph↗

arXiv2025

MXene triggers high toughness, high strength and low hysteresis hydrogels for printed artificial tissue

Substituting load-bearing tissues requires hydrogels with rapid processability, excellent mechanical strength and fatigue resistance. Conventional homogeneously polymerized hydrogels with short-chains/excessive branching exhibit low strength/toughness, being inadequate for artificial tissues. Here we introduce the heterogeneous polymerization-accelerated reaction kinetics on the Ti3C2Tx MXene microreactor and sluggish kinetics beyond-to rapidly produce hydrogels within minutes. This allows the hyperbranched domains embedded within a highly entangled matrix, leading to excellent strength (2.4 MPa)/toughness (75.2 kJ m-2) and low hysteresis (2.9%) in hydrogels superior to the rest ones. The rapid liquid-to-solid transition triggered by MXene suggests the great possibility of 3D printed robust hydrogels toward artificial tissue. Importantly, these printed hydrogels-based artificial ligaments have demonstrated impressive load-bearing capacity, wear resistance, and suturability compared to commercial analogs.

physics.chem-ph↗

arXiv2025

Deep Reparameterization for Full Waveform Inversion: Architecture Benchmarking, Robust Inversion, and Multiphysics Extension

Full waveform inversion (FWI) is a high-resolution subsurface imaging technique, but its effectiveness is limited by challenges such as noise contamination, sparse acquisition, and artifacts from multiparameter coupling. To address these limitations, this study develops a deep reparameterized FWI (DR-FWI) framework, in which subsurface parameters are represented by a deep neural network. Instead of directly optimizing the parameters, DR-FWI optimizes the network weights to reconstruct them, thereby embedding structural priors and facilitating optimization. To provide benchmark guidelines for the design of DR-FWI, we conduct a comparative analysis of three representative architectures (U-Net, CNN, MLP) combined with two initial model embedding strategies: one pretraining the network to generate predefined initial models (pretraining-based), while the other directly adds network outputs to the initial models. Extensive ablation experiments show that combining CNN with pretraining-based initialization significantly enhances inversion accuracy, offering valuable insights into network design. To further understand the mechanism of DR-FWI, spectral bias analysis reveals that the network first captures low-frequency features and gradually reconstructs high-frequency details, enabling an adaptive multi-scale inversion strategy. Notably, the robustness of DR-FWI is validated under various noise levels and sparse acquisition scenarios, where its strong performance with limited shots and receivers demonstrates reduced reliance on dense observational data. Additionally, a backbone-branch structure is proposed to extend DR-FWI to multiparameter inversion, and its efficacy in mitigating cross-parameter interference is validated on a synthetic anomaly model and the Marmousi2 model. These results suggest a promising direction for joint inversion involving multiple parameters or multiphysics.

physics.geo-ph↗

arXiv2024

Small-scale dynamics and structure of free-surface turbulence

The dynamics of small-scale structures in free-surface turbulence is crucial to large-scale phenomena in natural and industrial environments. Here we conduct experiments on the quasi-flat free surface of a zero-mean-flow turbulent water tank over the Reynolds number range $Re_{\lambda} = 207\textrm{--}312$. By seeding microscopic floating particles at high concentrations, the fine scales of the flow and the velocity gradient tensor are resolved. A kinematic relation is derived expressing the contribution of surface divergence and vorticity to the dissipation rate. The probability density functions of divergence, vorticity and strain-rate collapse once normalized by the Kolmogorov scales. Their magnitude displays strong intermittency and follows chi-square distributions with power-law tails at small values. The topology of high-intensity events and two-point statistics indicate that the surface divergence is characterized by dissipative spatial and temporal scales, while the high-vorticity and high-strain-rate regions are larger, long-lived, concurrent, and elongated. The second-order velocity structure functions obey the classic Kolmogorov scaling in the inertial range when the dissipation rate on the surface is considered, with a different numerical constant than in 3D turbulence. The cross-correlation among divergence, vorticity and strain-rate indicates that the surface-attached vortices are strengthened during downwellings and diffuse when those dissipate. Sources (sinks) in the surface velocity fields are associated with strong (weak) surface-parallel stretching and compression along perpendicular directions. The floating particles cluster over spatial and temporal scales larger than those of the sinks. These results demonstrate that, compared to 3D turbulence, in free-surface turbulence the energetic scales leave a stronger imprint on the small-scale quantities.

physics.flu-dyn↗

arXiv2024

A foundation model enpowered by a multi-modal prompt engine for universal seismic geobody interpretation across surveys

Seismic geobody interpretation is crucial for structural geology studies and various engineering applications. Existing deep learning methods show promise but lack support for multi-modal inputs and struggle to generalize to different geobody types or surveys. We introduce a promptable foundation model for interpreting any geobodies across seismic surveys. This model integrates a pre-trained vision foundation model (VFM) with a sophisticated multi-modal prompt engine. The VFM, pre-trained on massive natural images and fine-tuned on seismic data, provides robust feature extraction for cross-survey generalization. The prompt engine incorporates multi-modal prior information to iteratively refine geobody delineation. Extensive experiments demonstrate the model's superior accuracy, scalability from 2D to 3D, and generalizability to various geobody types, including those unseen during training. To our knowledge, this is the first highly scalable and versatile multi-modal foundation model capable of interpreting any geobodies across surveys while supporting real-time interactions. Our approach establishes a new paradigm for geoscientific data interpretation, with broad potential for transfer to other tasks.

physics.geo-ph↗

arXiv2024

Neuro-Inspired Information-Theoretic Hierarchical Perception for Multimodal Learning

Integrating and processing information from various sources or modalities are critical for obtaining a comprehensive and accurate perception of the real world in autonomous systems and cyber-physical systems. Drawing inspiration from neuroscience, we develop the Information-Theoretic Hierarchical Perception (ITHP) model, which utilizes the concept of information bottleneck. Different from most traditional fusion models that incorporate all modalities identically in neural networks, our model designates a prime modality and regards the remaining modalities as detectors in the information pathway, serving to distill the flow of information. Our proposed perception model focuses on constructing an effective and compact information flow by achieving a balance between the minimization of mutual information between the latent state and the input modal state, and the maximization of mutual information between the latent states and the remaining modal states. This approach leads to compact latent state representations that retain relevant information while minimizing redundancy, thereby substantially enhancing the performance of multimodal representation learning. Experimental evaluations on the MUStARD, CMU-MOSI, and CMU-MOSEI datasets demonstrate that our model consistently distills crucial information in multimodal learning scenarios, outperforming state-of-the-art benchmarks. Remarkably, on the CMU-MOSI dataset, ITHP surpasses human-level performance in the multimodal sentiment binary classification task across all evaluation metrics (i.e., Binary Accuracy, F1 Score, Mean Absolute Error, and Pearson Correlation).

cs.LG↗

arXiv2024

Neuron-based Multifractal Analysis of Neuron Interaction Dynamics in Large Models

In recent years, there has been increasing attention on the capabilities of large models, particularly in handling complex tasks that small-scale models are unable to perform. Notably, large language models (LLMs) have demonstrated ``intelligent'' abilities such as complex reasoning and abstract language comprehension, reflecting cognitive-like behaviors. However, current research on emergent abilities in large models predominantly focuses on the relationship between model performance and size, leaving a significant gap in the systematic quantitative analysis of the internal structures and mechanisms driving these emergent abilities. Drawing inspiration from neuroscience research on brain network structure and self-organization, we propose (i) a general network representation of large models, (ii) a new analytical framework, called Neuron-based Multifractal Analysis (NeuroMFA), for structural analysis, and (iii) a novel structure-based metric as a proxy for emergent abilities of large models. By linking structural features to the capabilities of large models, NeuroMFA provides a quantitative framework for analyzing emergent phenomena in large models. Our experiments show that the proposed method yields a comprehensive measure of network's evolving heterogeneity and organization, offering theoretical foundations and a new perspective for investigating emergent abilities in large models.

cs.AI↗

arXiv2023

Neuro-Inspired Hierarchical Multimodal Learning

Integrating and processing information from various sources or modalities are critical for obtaining a comprehensive and accurate perception of the real world. Drawing inspiration from neuroscience, we develop the Information-Theoretic Hierarchical Perception (ITHP) model, which utilizes the concept of information bottleneck. Distinct from most traditional fusion models that aim to incorporate all modalities as input, our model designates the prime modality as input, while the remaining modalities act as detectors in the information pathway. Our proposed perception model focuses on constructing an effective and compact information flow by achieving a balance between the minimization of mutual information between the latent state and the input modal state, and the maximization of mutual information between the latent states and the remaining modal states. This approach leads to compact latent state representations that retain relevant information while minimizing redundancy, thereby substantially enhancing the performance of downstream tasks. Experimental evaluations on both the MUStARD and CMU-MOSI datasets demonstrate that our model consistently distills crucial information in multimodal learning scenarios, outperforming state-of-the-art benchmarks.

cs.LG↗

arXiv2022

Evaporation of a sessile colloidal water-glycerol droplet: Marangoni ring formation

The transport and aggregation of particles in suspensions is an important process in many physicochemical and industrial processes. In this work, we study the transport of particles in an evaporating binary droplet. Surprisingly, the accumulation of particles occurs not only at the contact line (due to the coffee-stain effect) or at the solid substrate (due to sedimentation), but also at a particular radial position near the liquid-air interface, forming a "ring", which we term as the Marangoni ring. The formation of this ring is primarily attributed to the solutal Marangoni flow triggered by the evaporation dynamics of the water-glycerol droplet. Experiments and simulations show fair agreement in the volume evolution and the general structure of the solutal Marangoni flow, that is, the Marangoni vortex. Experiments show that the location of the Marangoni ring is strongly correlated with the Marangoni vortex. However, finite element numerical simulations fail to describe the particle distribution seen in the experiments. Interestingly, the particles not only accumulate to form the Marangoni ring but also assemble as colloidal crystals close to the liquid-air interface, yielding iridescence. The formation of the colloidal crystals in the experiments is strong evidence that non-hydrodynamic interactions, which are not represented in the simulations, also play a significant role in our system.

physics.flu-dyn↗

arXiv2022

Physiochemical hydrodynamics of the phase segregation in an evaporating binary microdroplet

Phase segregation triggered by selective evaporation can emerge in multicomponent systems, leading to complex physiochemical hydrodynamics. Recently, Li et al. (Phys. Rev. Lett., vol. 120, 2018, 224501) and Kim & Stone (J. Fluid Mech., vol. 850, 2018, pp. 769-783) reported a segregative behavior (i.e., demixing) in an evaporating binary droplet. In this work, by means of experiments and theoretical analysis, we investigate the flow dynamics after the occurrence of the phase segregation. As example, we take the 1,2-hexanediol-water binary droplet system. First, we experimentally reveal the overall physiochemical hydrodynamics of the evaporation process, including the segregative behavior and the resulting flow structure close to the substrate. By quantifying the evolution of the radial flow, we identify three successive life stages of the evaporation process. At Stage I, a radially outward flow is observed. It is driven by the Marangoni effect. At the transition to Stage II, the radial flow partially reverses, starting from the contact line. This flow breaks the axial symmetry and remarkably is driven by the segregation itself. Finally at Stage III, the flow decays as the evaporation gradually ceases. At this stage the segregation has grown to the entire droplet, and the flow is again controlled by the Marangoni effect. The resulting Marangoni flow homogenizes the distribution of the entrapped volatile water over the whole droplet.

physics.flu-dyn↗

arXiv2020

Evaporation-Induced Crystallization of Surfactants in Sessile Multicomponent Droplets

Surfactants have been widely studied and used in controlling droplet evaporation. In this work, we observe and study the crystallization of sodium dodecyl sulfate (SDS) within an evaporating glycerol-water mixture droplet. The crystallization is induced by the preferential evaporation of water, which decreases the solubility of SDS in the mixture. As a consequence, the crystals shield the droplet surface and cease the evaporation. The universality of the evaporation characteristics for a range of droplet sizes is revealed by applying a diffusion model, extended by Raoult's law. To describe the nucleation and growth of the crystals, we employ the 2-dimensional crystallization model of Weinberg [J. Non-Cryst. Solids 1991, 134, 116]. The results of this model compare favorably to our experimental results. Our findings may inspire the community to reconsider the role of high concentration of surfactants in a multicomponent evaporation system.

physics.flu-dyn↗

arXiv2020

Competing Marangoni and Rayleigh convection in evaporating binary droplets

For a small sessile or pendant droplet it is generally assumed that gravity does not play any role once the Bond number is small. This is even assumed for evaporating binary sessile or pendant droplets, in which convective flows can be driven due to selective evaporation of one component and the resulting concentration and thus surface tension differences at the air-liquid interface. However, recent studies have shown that in such droplets gravity indeed can play a role and that natural convection can be the dominant driving mechanism for the flow inside evaporating binary droplets (Edwards et al., Phys. Rev. Lett. 121, 184501 (2018); Li et al., Phys. Rev. Lett. 122, 114501 (2019)). In this study, we derive and validate a quasi-stationary model for the flow inside evaporating binary sessile and pendant droplets, which successfully allows to predict the prevalence and the intriguing interaction of Rayleigh and/or Marangoni convection on the basis of a phase diagram for the flow field expressed in terms of the Rayleigh and Marangoni numbers.

physics.flu-dyn↗

arXiv2020

Rayleigh-Taylor instability by segregation in an evaporating multi-component microdroplet

The evaporation of multi-component droplets is relevant to various applications but challenging to study due to the complex physicochemical dynamics. Recently, Li (2018) reported evaporation-triggered segregation in 1,2-hexanediol-water binary droplets. In this present work, we added 0.5 wt% silicone oil into the 1,2-hexanediol-water binary solution. This minute silicone oil concentration dramatically modifies the evaporation process as it triggers an early extraction of the 1,2-hexanediol from the mixture. Surprisingly, we observe that the segregation of 1,2-hexanediol forms plumes, rising up from the rim of the sessile droplet towards the apex during the droplet evaporation. By orientating the droplet upside down, i.e., by studying a pendant droplet, the absence of the plumes indicates that the flow structure is induced by buoyancy, which drives a Rayleigh-Taylor instability (i.e., driven by density differences & gravitational acceleration). From micro-PIV measurement, we further prove that the segregation of the non-volatile component (1,2-hexanediol) hinders the evaporation near the contact line, which leads to a suppression of the Marangoni flow in this region. Hence, on long time scales, gravitational effects play the dominant role in the flow structure, rather than Marangoni flows. We compare the measurement of the evaporation rate with the diffusion model of Popov (2005), coupled with Raoult's law and the activity coefficient. This comparison indeed confirms that the silicone-oil-triggered segregation of the non-volatile 1,2-hexanediol significantly delays the evaporation. With an extended diffusion model, in which the influence of the segregation has been implemented, the evaporation can be well described.

physics.flu-dyn↗

arXiv2019

Gravitational effect in evaporating binary microdroplets

The flow in an evaporating glycerol-water binary sub-millimeter droplet with Bond number Bo $\ll$ 1 is studied both experimentally and numerically. First, we measure the flow fields near the substrate by micro-PIV for both sessile and pendant droplets during the evaporation process, which surprisingly show opposite radial flow directions -- inward and outward, respectively. This observation clearly reveals that in spite of the small droplet size, gravitational effects play a crucial role in controlling the flow fields in the evaporating droplets. We theoretically analyze that this gravity-driven effect is triggered by the lower volatility of glycerol which leads to a preferential evaporation of water then the local concentration difference of the two components leads to a density gradient that drives the convective flow. We show that the Archimedes number Ar is the nondimensional control parameter for the occurrence of the gravitational effects. We confirm our hypothesis by experimentally comparing two evaporating microdroplet systems, namely a glycerol-water droplet and a 1,2-propanediol-water droplet. We obtain different Ar, larger or smaller than a unit by varying a series of droplet heights, which corresponds to cases with or without gravitational effects, respectively. Finally, we simulate the process numerically, finding good agreement with the experimental results and again confirming our interpretation.

physics.flu-dyn↗

arXiv2018

Evaporation-triggered segregation of sessile binary droplets

Droplet evaporation of multicomponent droplets is essential for various physiochemical applications, e.g. in inkjet printing, spray cooling and microfabrication. In this work, we observe and study phase segregation of an evaporating sessile binary droplet, consisting of a mixture of water and a surfactant-like liquid (1,2-hexanediol). The phase segregation (i.e., demixing) leads to a reduced water evaporation rate of the droplet and eventually the evaporation process ceases due to shielding of the water by the non-volatile 1,2-hexanediol. Visualizations of the flow field by particle image velocimetry and numerical simulations reveal that the timescale of water evaporation at the droplet rim is faster than that of the Marangoni flow, which originates from the surface tension difference between water and 1,2-hexanediol, eventually leading to segregation.

physics.flu-dyn↗