A5063383250- Enhanced Oil Recovery Techniques
- Hydraulic Fracturing and Reservoir Analysis
- Simulation Techniques and Applications
- Advanced Queuing Theory Analysis
- Hydrocarbon exploration and reservoir analysis
- Lattice Boltzmann Simulation Studies
- Drilling and Well Engineering
- Heat and Mass Transfer in Porous Media
- Machine Learning in Healthcare
- Groundwater flow and contamination studies
- Model Reduction and Neural Networks
- Generative Adversarial Networks and Image Synthesis
- Explainable Artificial Intelligence (XAI)
- Bayesian Modeling and Causal Inference
- Network Traffic and Congestion Control
- Markov Chains and Monte Carlo Methods
- Gaussian Processes and Bayesian Inference
- Cardiovascular Health and Disease Prevention
- Embedded Systems Design Techniques
- NMR spectroscopy and applications
- Rock Mechanics and Modeling
- Machine Learning and Algorithms
- Neural Networks and Applications
- Geophysical Methods and Applications
- Advanced Numerical Methods in Computational Mathematics
China University of Petroleum, Beijing
2024-2025
Tsinghua University
2022-2024
University of Chinese Academy of Sciences
2021-2023
Institute of Electrical Engineering
2021-2023
Chinese Academy of Sciences
2021-2023
Harbin Institute of Technology
2023
China University of Mining and Technology
2015-2022
University of Cambridge
2018-2021
Shandong University of Traditional Chinese Medicine
2021
University of Science and Technology Beijing
2020
Understanding and controlling fluid entrapment during forced imbibition in porous media is crucial for many natural industrial applications. However, the microscale physics macroscopic consequences of these geometric-confined remain poorly understood. Here, we introduce a novel multidepth microfluidic chip, which can mitigate depth confinement traditional two-dimensional (2-D) chips mimic wide pore size distribution as natural-occurring three-dimensional (3-D) media. Based on experiments...
We present a comprehensive description of the aspect ratio impact on interfacial instability in porous media where wetting liquid displaces nonwetting fluid. Building microfluidic experiments, we evidence imbibition scenarios yielding instabilities and macroscopic morphologies under different depth confinements, which were controlled by capillary number. report phenomenon whereby smaller depth-variable lower number trigger during forced imbibition; otherwise, larger uniform-depth higher will...
A gradient-estimation procedure for a general class of stochastic discrete-event systems is developed. In contrast to most previous work, the authors focus on performance measures whose realizations are inherently discontinuous (in fact, piecewise constant) functions parameter differentiation. Two broad classes finite-horizon arising naturally in applications considered. Because their discontinuity, these important not susceptible infinitesimal perturbation analysis (IPA). Instead, apply...
We assess the ability of large language models (LLMs) to answer causal questions by analyzing their strengths and weaknesses against three types question. believe that current LLMs can with existing knowledge as combined domain experts. However, they are not yet able provide satisfactory answers for discovering new or high-stakes decision-making tasks high precision. discuss possible future directions opportunities, such enabling explicit implicit modules well deep causal-aware LLMs. These...
Preferential flow in a porous medium is commonly encountered many practical applications. Our previous studies have discovered the preferential flow-induced non-monotonic wettability effect on displacement (J. Fluid Mech, vol. 942, 2022a, R5), but whether this rule consistent for different disordered media and impact of interplay between disorder under conditions still not well understood. Here, we combine microfluidic experiments, pore-scale simulations theoretical analysis to study...
An iterative discrete optimization algorithm that works with Monte Carlo estimation of the objective function is developed. Two algorithms, simulated annealing and stochastic ruler algorithm, are considered. The authors examine some problems their use combine advantages both algorithms to form an random search called comparison (SC) algorithm. SC actually solves alternative problem, it shown under symmetry assumption problem equivalent original one. convergence proved based on...
Utilizing the discrete element method and pore network model, we numerically investigate impact of compaction on longitudinal dispersion coefficient porous media. Notably, exhibits a non-monotonic dependence degree compaction, which is distinguished by presence three distinct regimes in variation coefficient. The attributed to disparate effect mechanisms. Specifically, medium tightens with an increasing pressure load, reducing molecular diffusion that primarily governs at small Péclet...
Training large language models (LLMs) typically relies on adaptive optimizers like Adam (Kingma & Ba, 2015) which store additional state information to accelerate convergence but incur significant memory overhead. Recent efforts, such as SWAN (Ma et al., 2024) address this by eliminating the need for optimizer states while achieving performance comparable via a multi-step preprocessing procedure applied instantaneous gradients. Motivated success of SWAN, we introduce novel framework...
Designing efficient optimizers for large language models (LLMs) with low-memory requirements and fast convergence is an important challenging problem. This paper makes a step towards the systematic design of such through lens structured Fisher information matrix (FIM) approximation. We show that many state-of-the-art can be viewed as solutions to FIM approximation (under Frobenius norm) specific structural assumptions. Building on these insights, we propose two recommendations practical...
Understanding the microscopic characteristics and evolutionary patterns of pore structures during high-PV waterflooding is critical for improving accuracy efficiency oil field development. While previous studies have primarily emphasized geometric morphological features overall structures, they often overlook local pore-scale properties their relationship with fluid transport capacity. This study proposes a novel classification method that integrates both size flow conductivity, enabling...
Abstract Accurate understanding and predicting the flow paths of immiscible two-phase in rocky porous structures are critical importance for evaluation oil or gas recovery prediction rock slides caused by gas-liquid flow. A 2D phase field model was established compressible air-water heterogenous structures. The dynamic characteristics interface preferential were simulated. factors affecting path selection analyzed. Transparent physical models complex prepared using 3D printing technology....
Abstract The pore network is an approximate representation of the void space porous materials, such as rocks and soil, via pores (corresponding to large cavities) throats (narrow constrictions). During extraction networks from real space, ambiguous definitions or determinations may cause significant errors in prediction single/multi‐phase transport properties. Meanwhile, pore‐throat segmentation needs exclude non‐physical parameters much possible. In this work, we propose a method based on...
To investigate the relationship between structural characteristics and seepage flow behavior of rough single rock fractures, a set fracture physical models were produced using Weierstrass–Mandelbrot functions to test performance. Six with various surface roughnesses characterized by fractal dimensions, built COMSOL multiphysics software. The fluid through fractures influences surfaces on was then monitored. numerical simulation indicates that there is linear average velocity over entire path...