A5024989765- Fluid Dynamics and Turbulent Flows
- Model Reduction and Neural Networks
- Fluid Dynamics and Vibration Analysis
- Biomimetic flight and propulsion mechanisms
- Wind and Air Flow Studies
- Computational Fluid Dynamics and Aerodynamics
- Lattice Boltzmann Simulation Studies
- Meteorological Phenomena and Simulations
- Vibration and Dynamic Analysis
- Plasma and Flow Control in Aerodynamics
- Magnetic and Electromagnetic Effects
- Structural Health Monitoring Techniques
- Solar Radiation and Photovoltaics
- Solar and Space Plasma Dynamics
- Hydrology and Sediment Transport Processes
- Fluid Dynamics and Mixing
- Planetary Science and Exploration
- Adaptive Control of Nonlinear Systems
- Aerosol Filtration and Electrostatic Precipitation
- Fluid Dynamics and Heat Transfer
- Robotic Path Planning Algorithms
- Radiative Heat Transfer Studies
- Nuclear Engineering Thermal-Hydraulics
- Aerospace Engineering and Energy Systems
- Combustion and flame dynamics
Zhejiang University
2017-2024
Nanjing University of Aeronautics and Astronautics
2017-2024
Massachusetts Institute of Technology
2015-2017
In the present work, an efficient active flow control strategy in eliminating vortex-induced vibration of a cylinder at Re = 100 has been explored by two machine learning frameworks, from to reinforcement learning. Specifically, adaptive scheme pair jets placed on poles as actuators discovered. framework, Gaussian progress regression surrogate model is used predict amplitude using limited number numerical simulations combining Bayesian optimization algorithm with specified actions while soft...
A novel data-driven nonlinear reduced-order modeling framework is proposed for unsteady fluid–structure interactions (FSIs). In the framework, a convolutional variational autoencoder model developed to determine coordinate transformation from high-dimensional physical field into reduced space. This enables efficient extraction of low-dimensional manifolds flow FSIs. The sparse identification dynamics (SINDy) algorithm then used identify dynamical governing equations space and vibration...
In the present work, we propose an optimization framework based on active learning method, which aims to quickly determine conditions of tandem flapping wings for optimal performance in terms thrust or efficiency. Especially, multi-fidelity Gaussian process regression is used establish surrogate model correlating kinematic parameters and their aerodynamic performances. Moreover, Bayesian algorithm employed select new candidate points update model. With this framework, parameter space can be...
A deep learning framework is proposed for real-time transonic flow prediction. To capture the complex shock discontinuity of flow, we introduce residual network ResNet and deconvolutional neural networks to learn nonlinear phenomenon in which affected by Mach number, angle attack, Reynolds aerodynamic shape. In our framework, field variables on actual grid points are utilized training avoid interpolation operation input spatial position with a point cloud that required traditional...
High-speed aircraft experiences severe aerodynamic heating at high Mach numbers, requiring accurate prediction of aerothermal effects before designing thermal protection systems. With the rise artificial intelligence and potential neural networks, data-driven methods for have gained significant attention. This study focuses on numerical simulations phenomena explores machine learning applications in heat prediction. First, a two-dimensional cylinder case was simulated using finite volume...
We employ three-dimensional direct and large-eddy numerical simulations of the vibrations flow past cylinders fitted with free-to-rotate U-shaped fairings placed in a cross-flow at Reynolds number $100\leqslant \mathit{Re}\leqslant 10\,000$ . Such are nearly neutrally buoyant devices along axis long circular risers to suppress vortex-induced (VIVs). consider three different geometric configurations: homogeneous fairing, two configurations (denoted A AB) involving gap between adjacent...
We study the instability mechanisms leading to slug flow formation in an inclined pipe subject gravity forces. use a phase-field approach, where Cahn–Hillard model is used interface. At inlet, stratified imposed with specified velocity profile. validate our numerical results by comparing against previous theoretical models and predicting various regimes for horizontal pipes, including flow, dispersed bubble annular flow. Subsequently, we focus on connect its appearance specific vortical...
Deep reinforcement learning (RL) is capable of identifying and modifying strategies for active flow control. However, the classic formulation deep RL requires lengthy exploration. This paper describes introduction expert demonstration into a off-policy algorithm, soft actor-critic application to vortex-induced vibration problems. combined online-learning framework applied an oscillator wake environment Navier–Stokes with obtained from pole-placement method surrogate model optimization. The...
In the present study, we use dynamic mesh method based on radial basis function interpolation for two-dimensional simulation of harmonically oscillating NACA0015 airfoil. Under various flapping frequencies, heaving and pitching amplitudes, observed wake flows can be divided into seven types, including B\'enard-von K\'arm\'an (BvK) vortex street, reversed BvK (RBvK) $1P$ wake, $mP$ $2P+mS$ $2S+mS$ $mS$ where $m$ is around 4 $xS+yP$ signifies $x$ single vortices $y$ pairs shedding per...
We construct a multifidelity framework for the kinematic parameter optimization of flapping airfoil. employ Gaussian process regression and Bayesian to effectively synthesize aerodynamic performance airfoil with parameters under multiresolution numerical simulations. The objective this work is demonstrate that can efficiently discover optimal specific using limited number expensive high-fidelity simulations combined larger inexpensive low-fidelity identify an asymmetrically various target...
In this paper, a control algorithm using multi-level optimization and model predictive is proposed to solve the conflict between computational cost accuracy of flapping-wing micro aerial vehicle. First, quasi-steady established evaluate aerodynamic forces moments vehicle, are then optimized meet requirements based on classical control. Then an module introduced optimize kinematic parameters achieve optimal moments, thus it decomposes complex problem into two sub-problems. Compared with...
This paper constructs an optimization framework based on the data-informed self-adaptive quasi-steady model. The aims at achieving a specific aerodynamic force coefficient by optimizing kinematic parameters of flapping motion ellipsoid wing. All model coefficients this are calibrated empirically training. At each iteration, training is implemented local ridge regression, where initial samples extracted from simulation examples, and weight calculated compactly supported radial basis function...
The wind shear layer is a naturally formed airflow that enables the albatross to soar for six days at almost no cost. modeling and prediction of can be very helpful long-endurance flight (dynamic soaring), but existing studies usually ignore turbulence structures layers. In this paper, on leeward side ridge simulated by large eddy simulation (LES) method analyze structures. numerical simulation, three-dimensional (3D) elevation data mountain used as topography bottom synthesized turbulent...
We present fully resolved simulations of the flow–structure interaction in a flexible pipe conveying incompressible fluid. It is shown that Reynolds number plays significant role onset flutter for fluid-conveying modelled through classic garden-hose problem. investigate complex between structural and internal flow dynamics obtain phase diagram transition states as function three non-dimensional quantities: fluid-tension parameter, dimensionless fluid velocity number. find patterns inside...
A state-of-the-art large eddy simulation code has been developed to solve compressible flows in turbomachinery. The engineered with a high degree of scalability, enabling it effectively leverage the many-core architecture new Sunway system. consistent performance 115.8 DP-PFLOPs achieved on high-pressure turbine cascade consisting over 1.69 billion mesh elements and 865 Degree Freedoms (DOFs). By leveraging high-order unstructured solver its portability heterogeneous parallel systems, we...
This study proposes a new robust and accurate immersed boundary method for the immersion of solid bodies within fluid with Cartesian grid. The present introduces signed distance fields to recognize geometry contours, eliminating need Lagrangian points. To fully maximize advantages offered by fields, general integration kernel formulation is introduced into direct forcing replace conventional regularized delta function. With combination function, an interpolation along radial direction...
Abstract Centrifugal modelling is widely recognized as a valuable approach in various fields, including slope and high dam engineering, geotechnical earthquake deep-sea advanced material preparation research. Zhejiang University building the Hypergravity Interdisciplinary Experiment Facility (CHIEF), poised to become largest fastest hypergravity centrifuge worldwide. A comprehensive analysis of internal airflow characteristics imperative for effective design centrifuge, velocity distribution...
Precise estimation of the thermal updraft environment is important for effective exploration wind resources in long-endurance drones. Nevertheless, previous regression algorithms exhibit limitations accurately evaluating updrafts under new operating conditions, and traditional airborne measurement methods are constrained by narrow ranges sparse spatial sampling. This study addresses these challenges harnessing continuous temperature data acquired via infrared sensors. The proposed...
An intelligent wind tunnel using an active learning approach automates flow control experiments to discover the aerodynamic impact of sweeping jets on a swept wing. A Gaussian process regression model is established study jet actuator's performance at various attack and flap deflection angles. By selectively focusing most informative experiments, proposed framework was able predict 3,721 wing conditions from just 55 significantly reducing number required leading faster cost-effective...
Utilizing thermal updrafts shows potential for enabling long-endurance cruising of fixed-wing unmanned aerial vehicles without energy consumption. This article presents a novel online method based on sparse identification nonlinear dynamics (SINDy) approach to achievement sources in the atmosphere. Initially, algorithm is incorporated into upper-level planning system, interacting with lower-level controller. Then, experiments are conducted through software-in-the-loop simulations (SITL)...
Active flow control based on reinforcement learning has received much attention in recent years. Indeed, the requirement for substantial data trial-and-error policies posed a significant impediment to their practical application, which also serves as limiting factor training of cross-case agents. This study proposes an in-context active policy framework grounded data. A transformer-based improvement operator is set up model process causal sequence and autoregressively give actions with...
Bionic flapping wing vehicles have great potential for civil and defense applications due to their flexibility concealment at low Reynolds numbers. Since traditional flow field pattern recognition methods are difficult identify effective information from the measured local deduce state of moving body, this study uses an artificial intelligence method establish internal correlation between information. Specifically, a fully connected neural network is adopted recognize tandem wings' by using...