A5014873301- Model Reduction and Neural Networks
- Fluid Dynamics and Turbulent Flows
- Lattice Boltzmann Simulation Studies
- Advanced Vision and Imaging
- Nuclear Engineering Thermal-Hydraulics
- Advanced Image Processing Techniques
- Image Enhancement Techniques
- Anomaly Detection Techniques and Applications
- Fluid Dynamics and Vibration Analysis
- Meteorological Phenomena and Simulations
- Distributed Control Multi-Agent Systems
- Seismic Imaging and Inversion Techniques
- Advanced Control Systems Optimization
- Autonomous Vehicle Technology and Safety
- Advanced Neural Network Applications
- Hemodynamic Monitoring and Therapy
- Glaucoma and retinal disorders
- Adaptive Control of Nonlinear Systems
- Advanced Control Systems Design
- Cardiac Valve Diseases and Treatments
- Aerodynamics and Acoustics in Jet Flows
- Medical Image Segmentation Techniques
- Robotic Path Planning Algorithms
- Blood properties and coagulation
- Retinal Imaging and Analysis
Zhejiang University
2020-2025
Zhejiang University of Technology
2016-2025
State Key Laboratory of Industrial Control Technology
2017-2025
Brown University
2020-2022
John Brown University
2022
Nanjing University of Science and Technology
2005
Abstract Physics-informed neural networks (PINNs) have gained popularity across different engineering fields due to their effectiveness in solving realistic problems with noisy data and often partially missing physics. In PINNs, automatic differentiation is leveraged evaluate differential operators without discretization errors, a multitask learning problem defined order simultaneously fit observed while respecting the underlying governing laws of Here, we present applications PINNs various...
Simulating and predicting multiscale problems that couple multiple physics dynamics across many orders of spatiotemporal scales is a great challenge has not been investigated systematically by deep neural networks (DNNs). Herein, we develop framework based on operator regression, the so-called network (DeepONet), with long term objective to simplify modeling avoiding fragile time-consuming "hand-shaking" interface algorithms for stitching together heterogeneous descriptions phenomena. To...
Particle image velocimetry (PIV), as a common technology for analyzing the global flow motion from images, plays significant role in experimental fluid mechanics. In this article, we investigate deep learning-based techniques such estimation problem. The aim of novel technique is to extract 2-D velocity fields images efficiently and accurately. First, introduce convolutional neural network (CNN) called LiteFlowNet, which proposed end-to-end optical estimation. Enhanced configurations...
Significance Microfluidics is an important in vitro platform to gain insights into mechanics of blood flow and mechanisms pathophysiology human diseases. Extraction 3D fields microfluidics with dense cell suspensions remains a formidable challenge. We present artificial-intelligence velocimetry (AIV) as general determine microaneurysm-on-a-chip simulate microaneurysms patients diabetic retinopathy. AIV built on physics-informed neural networks that integrate seamlessly 2D images from...
Quantifying the flow of cerebrospinal fluid (CSF) is crucial for understanding brain waste clearance and nutrient delivery, as well edema in pathological conditions such stroke. However, existing vivo techniques are limited to sparse velocity measurements pial perivascular spaces (PVSs) or low-resolution from brain-wide imaging. Additionally, volume rate, pressure, shear stress variation PVSs essentially impossible measure vivo. Here, we show that artificial intelligence velocimetry (AIV)...
Particle image velocimetry (PIV) and particle tracking (PTV) are important flow visualization technologies for measuring global velocity fields in a non-intrusive manner. However, they limited by the spatial resolution of measurement, require further post-processing steps to refine fields. To this end, we employ deep learning method, physics-informed neural networks (PINNs), which can integrate sparse measurements from PIV or PTV with governing equations fluid network. A real experiment,...
Tomographic background oriented schlieren (Tomo-BOS) imaging measures density or temperature fields in 3D using multiple camera BOS projections, and is particularly useful for instantaneous flow visualizations of complex fluid dynamics problems. We propose a new method based on physics-informed neural networks (PINNs) to infer the full continuous velocity pressure from snapshots obtained by Tomo-BOS imaging. PINNs seamlessly integrate underlying physics observed visualization data, hence...
Microaneurysms (MAs) are one of the earliest clinically visible signs diabetic retinopathy (DR). MA leakage or rupture may precipitate local pathology in surrounding neural retina that impacts visual function. Thrombosis MAs affect their turnover time, an indicator associated with and anatomic outcomes eyes. In this work, we perform computational modeling blood flow microchannels containing various to investigate pathologies DR. The particle-based model employed study can explicitly...
ABSTRACT Accurate target detection in low‐light environments is crucial for unmanned aerial vehicles (UAVs) and autonomous driving applications. In this study, the authors introduce a real‐time multimodal fusion enhanced (RMF‐ED), novel framework designed to overcome limitations of detection. By leveraging complementary capabilities near‐infrared (NIR) cameras light ranging (LiDAR) sensors, RMF‐ED enhances performance. An advanced NIR generative adversarial network (NIR‐GAN) model was...
ABSTRACT With the booming development of logistics, manufacturing and warehousing fields, autonomous navigation intelligent obstacle avoidance technology automated guided vehicles (AGVs) has become focus scientific research. In this paper, an enhanced deep reinforcement learning (DRL) framework is proposed, aiming to empower AGVs with ability in unknown variable complex environment. To address problems time‐consuming training limited generalisation traditional DRL, we refine twin delayed...
Obtaining high-precision aerodynamics in the automotive industry relies on large-scale simulations with computational fluid dynamics, which are generally time-consuming and computationally expensive. Recent advances operator learning for partial differential equations offer promising improvements terms of efficiency. However, capturing intricate physical correlations from extensive varying geometries while balancing discretization costs remains a significant challenge. To address these...
Purpose: Accurate segmentation of microaneurysms (MAs) from adaptive optics scanning laser ophthalmoscopy (AOSLO) images is crucial for identifying MA morphologies and assessing the hemodynamics inside MAs. Herein, we introduce AOSLO-net to perform automatic AOSLO diabetic retinas. Method: composed a deep neural network based on UNet with pretrained EfficientNet as encoder. We have designed customized preprocessing postprocessing policies images, including generation multichannel de-noising,...