检索结果-内蒙古大学图书馆

Multi-Scale Feature Convolutional Neural Network for Particle Image Velocimetry

IEEE Transactions on Instrumentation and Measurement 2025年

作者： Gu, Shusheng Wang, Jingyu Li, Guanxiong Deng, Xiaogang Sichuan University College of Computer Science The National key laboratory of Fundamental Algorithms and Models for Engineering Simulation 610065 China Sichuan University School of Aeronautics and Astronautics The National key laboratory of Fundamental Algorithms and Models for Engineering Simulation 610065 China Sichuan University National key laboratory of Fundamental Algorithms and Models for Engineering Simulation 610065 China

Particle Image Velocimetry (PIV) is a crucial technique in experimental fluid dynamics for non-invasively measuring the velocity components of flow fields. Deep learning methods applied to PIV for velocity field estimation are primarily derived from computer vision. However, these methods often fail to account for the distinctive characteristics of fluid dynamics, such as the very small scales of objects and displacements relative to pixel resolution, high-density particle distributions, complex and irregular motion patterns. As demonstrated in this work, these approaches encounter three main challenges: (i) failure to accurately estimate the motion for very small particles and displacements;(ii) significant computational discrepancies with high-density particle distributions;and (iii) considerable inaccuracies in the analysis of complex flow fields. To address these issues, we propose a novel convolutional neural network, named Multi-Scale Feature Recurrent All-Pairs Field Transforms (MSF-RAFT). Our approach involves: (1) developing a highresolution multi-scale feature extraction network;(2) integrating an innovative residual block to improve the network;and (3) creating a multi-scale feature correction module that constructs a four-layer correlation pyramid. The proposed network has been validated on various datasets and experimental images. MSF-RAFT exhibits superior performance, particularly with high-density small particle images, achieving a significant reduction in estimation error and accurately capturing small vortex details. Comparative analyses with benchmark models demonstrate that MSF-RAFT achieves up to a 33% reduction in estimation error compared to previous best models on complex datasets and provides a flow field estimation that more accurately represents the fluid dynamics around the nacelle. Additionally, evaluations of model size, runtime, and memory consumption demonstrate that MSF-RAFT is highly efficient, making it well-suited to meet the indu

关键词： Flow fields

来源：评论

学校读者我要写书评

暂无评论

LLaVA-ReID: Selective Multi-image Questioner for Interactive Person Re-Identification

arXiv

引用

arXiv 2025年

作者： Lu, Yiding Yang, Mouxing Peng, Dezhong Hu, Peng Lin, Yijie Peng, Xi College of Computer Science Sichuan University China National Key Laboratory of Fundamental Algorithms and Models for Engineering Numerical Simulation Sichuan University China

Traditional text-based person ReID assumes that person descriptions from witnesses are complete and provided at once. However, in real-world scenarios, such descriptions are often partial or vague. To address this limitation, we introduce a new task called interactive person re-identification (Inter-ReID). Inter-ReID is a dialogue-based retrieval task that iteratively refines initial descriptions through ongoing interactions with the witnesses. To facilitate the study of this new task, we construct a dialogue dataset that incorporates multiple types of questions by decomposing fine-grained attributes of individuals. We further propose LLaVA-ReID, a question model that generates targeted questions based on visual and textual contexts to elicit additional details about the target person. Leveraging a looking-forward strategy, we prioritize the most informative questions as supervision during training. Experimental results on both Inter-ReID and text-based ReID benchmarks demonstrate that LLaVA-ReID significantly outperforms baselines. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

The Turbomaster Structured Grid Generation Framework Based on Multi-Block Topology and Elliptic Pde Optimizer

SSRN

引用

SSRN 2024年

作者： Huang, Longlong Wang, Nianhua Liu, Cheng Si, Songlin Liu, Dong Huang, Siyuan Wang, Jingyu National Key Laboratory of Engineering Numerical Simulation Basic Algorithms and Models Sichuan University Sichuan Chendu China

With the widespread application of numerical simulation in aeroengine, it is crucial to automatically and efficiently generate structured grids for rotating components. Therefore, a comprehensive structured grid automatic generation software called TurboMaster was established. Firstly, we present the design details, including the software architecture and grid generation methods. Secondly, we provide a detailed demonstration of the Geometry Module and the Grid Module, which encompass automatic geometric model generation, topological construction, 2D multi-block grid generation and optimization, and 3D volume grid generation. Finally, we conduct an in-depth comparison between TurboMaster and two other well-known commercial software packages. The experimental results indicate that TurboMaster offers an automated platform spanning from geometric modeling to grid generation, significantly enhancing efficiency in grid generation for rotating machinery. Additionally, it improves mesh quality in certain aspects, except for some discrepancies in handling complex models. Overall, this system provides important technical support for the application of numerical simulation technology in aviation engines. © 2024, The Authors. All rights reserved.

关键词： Rotating machinery

来源：评论

学校读者我要写书评

暂无评论

The Turbomaster Structured Grid Generation Framework Based on Multi-Block Topology and Elliptic Pde Optimizer

SSRN

引用

SSRN 2024年

关键词： Rotating machinery

来源：评论

学校读者我要写书评

暂无评论

A Fully Differentiable Gnn-Based Pde Solver and Applications to Poisson's and Navier-Stokes Equations

SSRN

引用

SSRN 2024年

作者： Li, Tianyu Zou, Yiye Zou, Shufan Chang, Xinghua Zhang, Laiping Deng, Xiaogang National Key Laboratory of Fundamental Algorithms and Models for Engineering Simulation Sichuan University Chengdu610207 China School of Computer Science Sichuan University Chengdu610065 China College of Aerospace Science and Engineering National University of Defense Technology Changsha410000 China Academy of Military Sciences Beijing100190 China

In this study, we present a novel computational framework that integrates the finite volume method with graph neural networks to address the challenges in Physics-Informed Neural Networks(PINNs). Our approach leverages the flexibility of graph neural networks to adapt to various types of two-dimensional unstructured grids, enhancing the model's applicability across different physical equations and boundary conditions. The core innovation lies in the development of an unsupervised training algorithm that utilizes GPU parallel computing to implement a fully differentiable finite volume method discretization process. This method includes differentiable integral and gradient reconstruction algorithms, enabling the model to directly solve partial-differential equations(PDEs) during training without the need for pre-computed data. Our results indicate that the GNN model trained using this method can learn to solve two types of PDEs with multiple boundary conditions and multiple source terms under various unstructured grid discretizations in just one training sessions. Moreover, the trained GNN model achieves convergence during inference testing with at most only 1/4 of the iterative steps compared to CFD software. © 2024, The Authors. All rights reserved.

关键词： Finite volume method

来源：评论

学校读者我要写书评

暂无评论

numerical Analysis of the Overall Parameters of a Stratospheric Airship Influenced by Solar Array Layout Optimization

SSRN

引用

SSRN 2024年

作者： Li, Guanxiong He, Cheng Wang, Jingyu National Key Laboratory of Fundamental Algorithms and Models for Engineering Simulation Sichuan University Chengdu610000 China Unmanned Air Vehicle Research Institute Nanjing University of Aeronautics and Astronautics Nanjing210016 China School of Aeronautics and Astronautics Sichuan University Chengdu610000 China

Solar power is a vital energy source for stratospheric airships, and the layout of solar cells significantly influences both their performance and the airship's parameters. A numerical discretization method is employed to model the performance of curved solar arrays. Based on principles of energy balance and force equilibrium, a model is formulated to calculate the airship’s parameters. Following an analysis of curved solar panels’ performance, the investigation proceeds to assess the effects of solar cell performance on the airship’s parameters. An optimization platform for solar array layout is developed, which yields the most efficient placement strategy under varying heading angles. Research indicates that there is a remarkable disparity in the calculated performance of solar panels between using flat solar cell models versus curved ones. Furthermore, solar cell performance has a substantial impact on the airship’s parameters. By optimizing the solar array layout, the area and weight of solar cells can be significantly reduced;specifically, the area coverage of solar cells is reduced by 23.3%. Additionally, under equivalent mission scenarios, this optimization leads to a 5.1% reduction in the airship's total weight, and, with the same total weight limitation, it boosts the airship's payload capacity by 1.58 times. © 2024, The Authors. All rights reserved.

关键词： Airships

来源：评论

学校读者我要写书评

暂无评论

A fully differentiable GNN-based PDE Solver: With Applications to Poisson and Navier-Stokes Equations

arXiv

引用

arXiv 2024年

In this study, we present a novel computational framework that integrates the finite volume method with graph neural networks to address the challenges in Physics-Informed Neural Networks(PINNs). Our approach leverages the flexibility of graph neural networks to adapt to various types of two-dimensional unstructured grids, enhancing the model's applicability across different physical equations and boundary conditions. The core innovation lies in the development of an unsupervised training algorithm that utilizes GPU parallel computing to implement a fully differentiable finite volume method discretization process. This method includes differentiable integral and gradient reconstruction algorithms, enabling the model to directly solve partial-differential equations(PDEs) during training without the need for pre-computed data. Our results demonstrate the model's superior mesh generalization and its capability to handle multiple boundary conditions simultaneously, significantly boosting its generalization capabilities. The proposed method not only shows potential for extensive applications in CFD but also establishes a new paradigm for integrating traditional numerical methods with deep learning technologies, offering a robust platform for solving complex physical problems. Copyright © 2024, The Authors. All rights reserved.

关键词： Finite volume method

来源：评论

学校读者我要写书评

暂无评论

Granular-ball fuzzy information-based outlier detector

引用

International Journal of Approximate Reasoning 2025年 185卷

作者： Li, Qilin Yuan, Zhong Peng, Dezhong Song, Xiaomin Zheng, Huiming Su, Xinyu Marketing Service Center of State Grid Sichuan Electric Power Corporation Chengdu610045 China College of Computer Science Sichuan University Chengdu610065 China Sichuan National Innovation New Vision UHD Video Technology Co. Ltd Chengdu610095 China Tianfu Jincheng Laboratory Chengdu610093 China National Key Laboratory of Fundamental Algorithms and Models for Engineering Numerical Simulation Chengdu610207 China

Outlier detection is an important part of the process of carrying out data mining and analysis and has been applied to many fields. Existing methods are typically anchored in a single-sample processing paradigm, where the processing unit is each individual and single-granularity sample. This processing paradigm is inefficient and ignores the multi-granularity features inherent in data. In addition, these methods often overlook the uncertainty information present in the data. To remedy the above-mentioned shortcomings, we propose an unsupervised outlier detection method based on Granular-Ball Fuzzy Granules (GBFG). GBFG adopts a granular-ball-based computing paradigm, where the fundamental processing units are granular-balls. This shift from individual samples to granular-balls enables GBFG to capture the overall data structure from a multi-granularity perspective and improve the performance of outlier detection. Subsequently, we calculate the outlier factor based on the outlier degrees of the granular-ball fuzzy granules to which the sample belongs, serving as a measure of the outlier degrees of samples. The experimental results prove that GBFG has a remarkable performance compared with the existing excellent algorithms. The code of GBFG is publicly available on https://***/Mxeron/GBFG. © 2025 Elsevier Inc.

关键词： Granular computing

来源：评论

学校读者我要写书评

暂无评论

Finite-difference-informed graph network for solving steady-state incompressible flows on block-structured grids

arXiv

引用

arXiv 2024年

作者： Zou, Yiye Li, Tianyu Lu, Lin Wang, Jingyu Zou, Shufan Zhang, Laiping Deng, Xiaogang School of Computer Science Sichuan University Chengdu610065 China National Key Laboratory of Fundamental Algorithms and Models for Engineering Simulation Sichuan University Chengdu610207 China School of Aeronautics and Astronautics Sichuan University Chengdu610065 China College of Aerospace Science and Engineering National University of Defense Technology Changsha410000 China Academy of Military Sciences Beijing100190 China

Advances in deep learning have enabled physics-informed neural networks to solve partial differential equations. numerical differentiation using the finite-difference (FD) method is efficient in physics-constrained designs, even in parameterized settings. In traditional computational fluid dynamics (CFD), body-fitted block-structured grids are often employed for complex flow cases when obtaining FD solutions. However, convolution operators in convolutional neural networks for FD are typically limited to single-block grids. To address this issue, graphs and graph networks are used to learn flow representations across multi-block-structured grids. A graph convolution-based FD method (GC-FDM) is proposed to train graph networks in a label-free physics-constrained manner, enabling differentiable FD operations on unstructured graph outputs. To demonstrate model performance from single- to multi-block-structured grids, the parameterized steady incompressible Navier–Stokes equations are solved for a lid-driven cavity flow and the flows around single and double circular cylinder configurations. When compared to a CFD solver under various boundary conditions, the proposed method achieves a relative error in velocity field predictions in the order of 10-3. Furthermore, the proposed method reduces training costs by approximately 20% compared to a physics-informed neural network. To further verify the effectiveness of GC-FDM in multi-block processing, a 30P30N airfoil geometry is considered, and the predicted results are reasonably compared with those given by CFD. Finally, the applicability of GC-FDM to a three-dimensional (3D) case is tested using a 3D cavity geometry. Copyright © 2024, The Authors. All rights reserved.

关键词： Computational fluid dynamics

来源：评论

学校读者我要写书评

暂无评论

Efficient neural network training method for unsteady flow field prediction based on data pool

引用

engineering Applications of Computational Fluid Mechanics 2025年第1期19卷

作者： Qiuchi Min Tianyu Li Guanxiong Li Shuyuan Liu Laiping Zhang Xiaogang Deng a School of Computer Science Sichuan University Chengdu People’s Republic of Chinab National Key Laboratory of Fundamental Algorithms and Models for Engineering Numerical Simulation Sichuan University Chengdu People’s Republic of China a School of Computer Science Sichuan University Chengdu People’s Republic of China b National Key Laboratory of Fundamental Algorithms and Models for Engineering Numerical Simulation Sichuan University Chengdu People’s Republic of China

In recent years, neural network technology has made significant progress in the field of unsteady flow field prediction, leading to the development of many innovative methods. However, deep learning-based unsteady flow field prediction techniques typically rely on autoregressive models, which inevitably face the issue of error accumulation. Existing solutions often suffer from challenges such as complex hyperparameter configurations and reduced training efficiency. To address these issues, this study makes the following contributions: (1) A novel training method is proposed to enhance model convergence by refining the training strategies employed. This approach achieves improved performance without necessitating complex hyperparameter conﬁgurations. (2) An innovative curriculum learning-based timestep reset strategy is introduced. This strategy further improves the convergence of neural networks and enhances prediction accuracy. A detailed comparative study was conducted on different training methods within the architecture of convolutional neural networks. Experimental results show that the proposed training strategy significantly improves prediction accuracy, with an improvement of up to an order of magnitude. Moreover, even when the training set spans only 600 timesteps, the model remains stable when predicting up to 9000 timesteps. Finally, our method also demonstrates high efficiency in terms of training time.

关键词： Neural network unsteady flow field prediction attention U-net data pool

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：