A5031719304- Computational Fluid Dynamics and Aerodynamics
- Model Reduction and Neural Networks
- Gas Dynamics and Kinetic Theory
- Gaussian Processes and Bayesian Inference
- Lattice Boltzmann Simulation Studies
- Advanced Numerical Methods in Computational Mathematics
- Reservoir Engineering and Simulation Methods
- Probabilistic and Robust Engineering Design
- Aerospace Engineering and Energy Systems
- Seismic Imaging and Inversion Techniques
- Meteorological Phenomena and Simulations
- Composite Material Mechanics
- Groundwater flow and contamination studies
- Neural Networks and Applications
- Fluid Dynamics Simulations and Interactions
- Tensor decomposition and applications
- Numerical methods for differential equations
- Advanced Numerical Analysis Techniques
- Bayesian Methods and Mixture Models
- Advanced Neuroimaging Techniques and Applications
- Urinary Tract Infections Management
- Random Matrices and Applications
- Numerical methods in engineering
- Advanced Multi-Objective Optimization Algorithms
- Markov Chains and Monte Carlo Methods
California Institute of Technology
2020-2025
Peking University
2023-2025
Peking University International Hospital
2024
San Jose State University
2021-2023
Stanford University
2018-2021
Software Engineering Institute
2019
Deep learning surrogate models have shown promise in solving partial differential equations (PDEs). Among them, the Fourier neural operator (FNO) achieves good accuracy, and is significantly faster compared to numerical solvers, on a variety of PDEs, such as fluid flows. However, FNO uses Fast transform (FFT), which limited rectangular domains with uniform grids. In this work, we propose new framework, viz., geo-FNO, solve PDEs arbitrary geometries. Geo-FNO learns deform input (physical)...
Bacteria can swim upstream in a narrow tube and pose clinical threat of urinary tract infection to patients implanted with catheters. Coatings structured surfaces have been proposed repel bacteria, but no such approach thoroughly addresses the contamination problem Here, on basis physical mechanism swimming, we propose novel geometric design, optimized by an artificial intelligence model. Using
Summary Embedded Boundary Methods (EBMs) are often preferred for the solution of Fluid‐Structure Interaction (FSI) problems because they reliable large structural motions/deformations and topological changes. For viscous flow problems, however, do not track boundary layers that form around embedded obstacles therefore maintain them resolved. Hence, an Adaptive Mesh Refinement (AMR) framework EBMs is proposed in this paper. It based on computing distance from edge embedding computational...
Abstract We consider Bayesian inference for large-scale inverse problems, where computational challenges arise from the need repeated evaluations of an expensive forward model. This renders most Markov chain Monte Carlo approaches infeasible, since they typically require <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" overflow="scroll"> <mml:mi mathvariant="script">O</mml:mi> <mml:mrow> <mml:mo stretchy="false">(</mml:mo> <mml:mn>1</mml:mn> <mml:msup>...
A high fidelity multi-physics Eulerian computational framework is presented for the simu- lation of supersonic parachute inflation during Mars landing. Unlike previous investigations in this area, takes into account an initial folding pattern parachute, flow compressibility effect on fabric material porosity, and interactions between super- sonic fluid flows suspension lines. Several adaptive mesh refinement (AMR)-enabled, large edge simulation (LES)-based, simulations a full-size...
SUMMARY Cable subsystems characterized by long, slender, and flexible structural elements are featured in numerous engineering systems. In each of them, interaction between an individual cable the surrounding fluid is inevitable. Such a fluid‐structure has received little attention literature, possibly due to inherent complexity associated with semidiscretizations disparate spatial dimensions. This article proposes embedded boundary approach for filling this gap, where dynamics captured...
EnsembleKalmanProcesses.jl is a Julia-based toolbox that can be used for broad class of black-box gradient-free optimization problems.Specifically, the tools enable optimization, or calibration, parameters within computer model in order to best match user-defined outputs with available observed data (Kennedy & O'Hagan, 2001).Some also approximately quantify parametric uncertainty (Huang, Huang, et al., 2022).Though package written Julia (Bezanson 2017), read-write TOML-file interface...
Abstract In this paper, we study efficient approximate sampling for probability distributions known up to normalization constants. We specifically focus on a problem class arising in Bayesian inference large-scale inverse problems science and engineering applications. The computational challenges address with the proposed methodology are: (i) need repeated evaluations of expensive forward models; (ii) potential existence multiple modes; (iii) fact that gradient of, or adjoint solver for,...
This paper is concerned with the approximation of probability distributions known up to normalization constants, a focus on Bayesian inference for large-scale inverse problems in scientific computing. In this context, key challenges include costly repeated evaluations forward models, multimodality, and inaccessible gradients model. To address them, we develop variational framework that combines Fisher-Rao natural gradient specialized quadrature rules enable derivative free updates Gaussian...
Surrogate models are critical for accelerating computationally expensive simulations in science and engineering, particularly solving parametric partial differential equations (PDEs). Key challenges developing practical surrogate include managing high-dimensional inputs outputs handling geometrically complex variable domains, which often represented as point clouds. In this work, we systematically investigate the formulation of neural operators on clouds introduce Point Cloud Neural Operator...
We develop a Koopman reduced-order model (ROM) to analyze the instability mechanism and predict hydrodynamic behavior for flag flapping in wake of cylinder. The ROM is constructed using kernel dynamic mode decomposition method enhanced through residual dynamical algorithm, which improves accuracy by identifying eliminating spurious modes. Our analysis reveals flow transition from "2S" periodic phase "2P" quasiperiodic phase, with main M_{1} providing insights into mechanism. In case chaotic...
View Video Presentation: https://doi.org/10.2514/6.2023-0370.vid The selection of the type and placement sensors is crucial to successfully predicting structural health a system. This paper expands on previous work involving sensor for simple beam apply same methodology more complex geometry. problem solved using machine learning feature importance selection, in this case random forest regression. features correspond signal measured by accelerometers or strain gauges, output corresponds...