检索结果-内蒙古大学图书馆

A computationally efficient hybrid neural network architecture for porous media: Integrating convolutional and graph neural networks for improved property predictions

引用

ADVANCES IN WATER RESOURCES 2025年 195卷

作者： Zhao, Qingqi Han, Xiaoxue Guo, Ruichang Chen, Cheng Univ Stuttgart Stuttgart Ctr Simulat Sci SC SimTech Stuttgart Germany Stevens Inst Technol Dept Comp Sci Hoboken NJ USA Univ Texas Austin Bur Econ Geol Austin TX USA Stevens Inst Technol Dept Civil Environm & Ocean Engn Hoboken NJ 07030 USA

Porous media is widely distributed in nature, found in environments such as soil, rock formations, and plant tissues, and is crucial in applications like subsurface oil and gas extraction, medical drug delivery, and filtration systems. Understanding the properties of porous media, such as the permeability and formation factor, is crucial for comprehending the physics of fluid flow within them. We present a novel fusion model that significantly enhances memory efficiency compared to traditional convolutional neural networks (CNNs) while maintaining high predictive accuracy. Although the CNNs have been employed to estimate these properties from highresolution, three-dimensional images of porous media, they often suffer from high memory consumption when processing large-dimensional inputs. Our model integrates a simplified CNN with a graph neural network (GNN), which efficiently consolidates clusters of pixels into graph nodes and edges that represent pores and throats, respectively. This graph-based approach aligns naturally with the porous medium structure, enabling large-scale simulations that are challenging with traditional methods. Furthermore, we use the GNN Grad-CAM technology to provide new interpretability and insights into fluid dynamics in porous media. Our results demonstrate that the accuracy of the fusion model in predicting porous medium properties is superior to that of the standalone CNN, while its total parameter count is nearly two orders of magnitude lower. This innovative approach highlights the transformative potential of hybrid neural network architectures in advancing research on fluid flow in porous media.

关键词： Deep learning Graph neural network Permeability Porous media Pore network modeling

来源：评论

学校读者我要写书评

暂无评论

Coupled Noise Reduction in distributed Acoustic Sensing Seismic Data Based on Convolutional neural network

引用

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS 2022年 19卷 1页

作者： Zhao, Yuxing Li, Yue Wu, Ning Jilin Univ Coll Commun Engn Dept Informat Changchun 130012 Peoples R China

distributed acoustic sensing (DAS) is widely recognized as a new technology to replace conventional geophones for the acquisition of seismic data. However, the collected data often contain a lot of coupled noise due to cable slapping and ringing along the borehole casing, which brings great difficulties to the interpretation of seismic data. The existing conventional coupled noise reduction methods often need to estimate the parameters of each coupled noise (such as amplitude, noise period, attenuation coefficient, etc.), which takes a lot of time and cannot meet the requirements for large-data-volume DAS seismic data processing. In addition, some deep learning-based denoising methods lack detailed analysis on coupled noise and have problems in the construction of training sets, resulting in insufficient generalization ability of the denoising model. To solve these problems, we propose a coupled noise reduction method based on the convolutional neural network (CNN). The proposed method does not need to estimate the parameters of coupled noise, and the denoising process is more convenient and efficient. In addition, through the analysis of DAS seismic data, we also construct a training set for coupled noise reduction using real data and synthetic data. The denoising results of both synthetic data and field data show that the proposed method can effectively reduce the coupled noise in DAS seismic data, and the effective signal has almost no energy loss. After processing, the signal affected by coupled noise becomes clear and continuous, providing high-quality data support for subsequent interpretation.

关键词： Noise reduction Training Attenuation Convolutional neural networks Data processing Mathematical models Geoscience and remote sensing Convolutional neural network (CNN) coupled noise denoise distributed acoustic sensing (DAS) training set

来源：评论

学校读者我要写书评

暂无评论

Flyback PPC Converter with Artificial neural network for distributed MPPT Application 8

Flyback PPC Converter with Artificial Neural Network for Dis...

引用

8th International Conference on Advanced Computing and Communication Systems, ICACCS 2022

作者： Sebin Davis, K. Jegathesan, V. Karunya University Department of Eee Coimbatore641114 India

ISBN: (纸本)9781665408165

This paper proposes a partial power processing (PPC) based configuration of DC-DC converter to overcome the issues due to partial shading in photovoltaic (PV) modules. Series connected partial power converter (S-PPC) with flyback topology is utilised for the implementation of distributed maximum power point tracking (DMPPT). In a conventional DC-DC converter, the total string power is processed by a single converter known as global maximum power point tracking (GMPPT) is utilised. Again, in the conventional DC-DC converters, the converter process complete power to achieve maximum power point tracking (MPPT). This paper proposes a Flyback S-PPC circuit which process only a part of total power flow, which is known as a partial power processing (PPP). The Flyback S-PPC is connected to each PV module and ANN-based MPPT is applied, higher efficiency with good steady-state and better transient responses are obtained. A comparative study is conducted to validate the performance of the proposed DC-DC converter configuration, and its results are discussed in this paper © 2022 IEEE.

关键词： DC-DC converters

来源：评论

学校读者我要写书评

暂无评论

Comprehensive study on hot deformation behaviour and mechanism of Ti-900 alloy: Insights from experimental study, neural network modeling and finite element simulation

引用

JOURNAL OF ALLOYS AND COMPOUNDS 2025年 1026卷

作者： Chakraborty, Dipayan Khan, Nisar Ahamad Sriman, V Reddy, B. R. Chandra Obul Kumar, Ajay Indian Inst Technol Dept Mech Engn Tirupati 517619 India Natl Inst Technol Dept Met & Mat Engn Warangal 506004 India DRDO Gas Turbine Res Estab Bangalore 560093 India

Thermomechanical processing plays a key role in shaping and refining the microstructure of structural components. This study examines the high-temperature deformation behaviour of equiaxed Ti-900, an alpha + (3 titanium alloy, through hot compression tests conducted at temperatures ranging from 850 degrees C to 1050 degrees C, strain rates from 0.001 s-1 to 10 s-1, and two different strain levels. An artificial neural network (ANN) model, optimized using a backpropagation algorithm, predicts flow stress with high accuracy. processing maps are developed to assess deformation efficiency at different reduction ratios. As strain increases, the stable deformation efficiency region becomes more confined. Flow stress increases with a higher reduction ratio, while high strain rates cause instability from flow localization and adiabatic shear, leading to crack formation. Q-Form simulations validate the deformation behaviour across different processing regions, aiding in the defect-free manufacturing of gas turbine compressor stator blades. A significant variation in effective stress is observed in samples deformed within the unstable regions identified from the processing map, whereas effective stress remains more uniformly distributed in stable regions. The deformation mechanism of the alloy is developed, highlighting the influence of temperature and strain rate on flow softening mechanisms such as dynamic recovery (DRV) and dynamic recrystallization (DRX). At a constant deformation temperature, lower strain rates promote a higher degree of DRX compared to higher strain rates. Deformation just below the (3 transus temperature favours continuous dynamic recrystallization (CDRX), while deformation above the (3 transus temperature facilitates DRX within the (3 phase.

关键词： Ti-900 alpha + beta titanium alloy Hot deformation Artificial neural network (ANN) processing map FE simulation Dynamic recovery Dynamic recrystallization

来源：评论

学校读者我要写书评

暂无评论

distributed data processing and task scheduling based on GPU parallel computing

引用

neural Computing and Applications 2024年第4期37卷 1757-1769页

作者： Li, Jun Department of Electronic Information Engineering Nanchong Vocational and Technical College Sichuan Nanchong637131 China

distributed data parallel (DDP) computing ensures data parallelism, enabling execution across several computers. A separate distributed data parallel computing instance should be created for each process that uses distributed data parallel computing. Task scheduling in parallel processing employs various methods and strategies to minimize the number of delayed jobs. Reliability in terms of a system’s capacity to distribute work over several machines and computers is improved by workload sharing, task migration, and automatic task replication;data transmission and reception are facilitated via the internet. The graphics processing unit (GPU) is a highly specialized electrical circuit. Compared to a traditional computer, a GPU parallel structure allows faster computation and greater effectiveness. GPUs are employed in parallel processing in new methods for DDP-GPUs with more processors, allowing a more significant workload to be handled in portions. The execution time of a program can be reduced by distributing various task areas among many processors. This study employs various multitasking scenarios, from simulated to real-world cases and from graphically heavy applications to parallel-processing workloads. The research compares the efficacy of several allocation algorithms, illuminating how best to divide GPU resources among multiple processes. By sorting through a vast volume of data more slowly, parallel computing saves time and money. Parallel and distributed computing concern the use of numerous computing resources to improve the results of a distributed and computationally expensive application. A single computer or a group of computers connected via a network can be used for computation. Our experimental findings show that the DPP-GPU implements task scheduling 95.3% better than the conventional GPU. The proposed model improves system reliability with an average execution time of 25.7 s. The time needed to run a program divided by the machine’s cost is approx

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

distributed photovoltaic cluster output monitoring method based on time series data acquisition

引用

Energy Informatics 2025年第1期8卷 1-27页

作者： Ye, Hua Lu, Xuegang Zhang, Wei Cheng, Fei Zhao, Ying School of Electrical and Information Engineering Tianjin University Tianjin300072 China Yunnan Power Grid Co. Ltd Yunnan Kunming650200 China

The data processing efficiency of distributed photovoltaic cluster output monitoring needs to be improved, improving the prediction effect of distributed photovoltaic power station cluster can effectively improve the security of power system operation and reduce the difficulty of power grid management. In order to obtain a reliable distributed photovoltaic cluster output monitoring method, this paper analyzes the output relationship of cluster power stations, combining time series data analysis methods for distributed cluster processing and monitoring data processing, a combined model of ceemdan and Bayesian neural network is proposed, the representative power plant prediction values obtained by establishing a combination model are weighted to obtain the cluster output prediction values. Compared with the simple superposition of the predicted values of cluster power stations, the average absolute error of this method is reduced by 3.3%, and the root mean square error is reduced by 5.86%. It is concluded that this model can effectively predict the power stations in the cluster. According to the experimental analysis, the output monitoring method of distributed photovoltaic clusters based on time series data collection proposed in this paper has certain effects and can provide theoretical support for the further development of distributed photovoltaic clusters. © The Author(s) 2025.

关键词： Data acquisition distributed photovoltaics Monitoring Time series

来源：评论

学校读者我要写书评

暂无评论

HiHGNN: Accelerating HGNNs Through Parallelism and Data Reusability Exploitation

引用

IEEE TRANSACTIONS ON PARALLEL AND distributed SYSTEMS 2024年第7期35卷 1122-1138页

作者： Xue, Runzhen Han, Dengke Yan, Mingyu Zou, Mo Yang, Xiaocheng Wang, Duo Li, Wenming Tang, Zhimin Kim, John Ye, Xiaochun Fan, Dongrui Chinese Acad Sci Inst Comp Technol State Key Lab Processors Beijing 100045 Peoples R China Univ Chinese Acad Sci Beijing 101408 Peoples R China Shanghai Innovat Ctr Processor Technol Shanghai 201203 Peoples R China Korea Adv Inst Sci & Technol Sch Elect Engn Daejeon 305701 South Korea

Heterogeneous graph neural networks (HGNNs) have emerged as powerful algorithms for processing heterogeneous graphs (HetGs), widely used in many critical fields. To capture both structural and semantic information in HetGs, HGNNs first aggregate the neighboring feature vectors for each vertex in each semantic graph and then fuse the aggregated results across all semantic graphs for each vertex. Unfortunately, existing graph neural network accelerators are ill-suited to accelerate HGNNs. This is because they fail to efficiently tackle the specific execution patterns and exploit the high-degree parallelism as well as data reusability inside and across the processing of semantic graphs in HGNNs. In this work, we first quantitatively characterize a set of representative HGNN models on GPU to disclose the execution bound of each stage, inter-semantic-graph parallelism, and inter-semantic-graph data reusability in HGNNs. Guided by our findings, we propose a high-performance HGNN accelerator, HiHGNN, to alleviate the execution bound and exploit the newfound parallelism and data reusability in HGNNs. Specifically, we first propose a bound-aware stage-fusion methodology that tailors to HGNN acceleration, to fuse and pipeline the execution stages being aware of their execution bounds. Second, we design an independency-aware parallel execution design to exploit the inter-semantic-graph parallelism. Finally, we present a similarity-aware execution scheduling to exploit the inter-semantic-graph data reusability. Compared to the state-of-the-art software framework running on NVIDIA GPU T4 and GPU A100, HiHGNN respectively achieves an average 40.0x and 8.3x speedup as well as 99.59% and 99.74% energy reduction with quintile the memory bandwidth of GPU A100.

关键词： Semantics Parallel processing Graph neural networks Vectors Graphics processing units Fuses Hardware GNN GNN accelerator graph neural network HGNN HGNN accelerator heterogeneous graph neural network

来源：评论

学校读者我要写书评

暂无评论

Leveraging Cloud Resources for distributed Training of Residual CNNs in Aerial Image Classification

引用

IEEE ACCESS 2025年 13卷 31131-31139页

作者： Kumar, Shantanu Singh, Shruti Kumar Dewangan, Narendra Amazon Seattle WA 98109 USA Washington State Univ IREACH Seattle WA 98195 USA Univ Petr & Energy Studies Sch Comp Sci Dehra Dun 248007 Uttaranchal India

Aerial image classification is crucial across multiple sectors, including environmental monitoring, agriculture, and urban planning. However, processing large-scale aerial imagery efficiently poses challenges in model performance, computational efficiency, and scalability. This research introduces a novel convolutional neural network (CNN) architecture tailored for cactus identification from aerial photographs. The proposed cloud-based pipeline enhances training efficiency through scalable data storage, preprocessing, and distributed training across platforms such as Amazon Web Services (AWS), Google Cloud Platform (GCP), and Microsoft Azure. This comparative analysis demonstrates the model's computational efficiency, shorter training durations, and cost-effectiveness. The model integrates residual connections and depthwise separable convolutions, achieving 96.7% accuracy on the aerial cactus identification dataset. The results highlight the model's high performance, cost-efficiency, and scalability, making it suitable for real-world aerial image classification tasks. Future work aims to further optimize the model using advanced techniques and extend its application to multi-class classification challenges.

关键词： Cloud computing Computational modeling Training Artificial intelligence Data models Image classification Computer architecture Computational efficiency Accuracy Scalability Aerial image classification convolutional neural network (CNN) cloud computing distributed model training

来源：评论

学校读者我要写书评

暂无评论

Parmesan: Efficient Partitioning and Mapping Flow for DNN Training on General Device Topology

引用

IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS 2024年第8期43卷 2426-2439页

作者： Liu, Lixin Liu, Tianji Jiang, Bentian Young, Evangeline F. Y. Chinese Univ Hong Kong Dept Comp Sci & Engn Hong Kong Peoples R China

Recently, various pipeline parallelism strategies are proposed to tackle the scalability problem of training a large DNN model on a distributed system. However, most of the works focus on pipeline scheduling while lacking a general methodology to handle network partitioning and mapping to distributed systems with heterogeneous interconnection. In this work, we propose an efficient design flow, named Parmesan, to map the training of a large DNN onto a system with general device topology to maximize the throughput. Parmesan works in an end-to-end manner and solves the whole optimization problem in two phases. The first phase aims at producing well-balanced partitions, and the second phase works toward placing the DNN on devices connected by an arbitrary topology network, considering the heterogeneity of the interconnection bandwidth. We show that Parmesan speeds up the pipeline training throughput on systems with different GPU topologies and is able to handle the mapping problem for heterogeneously interconnected architectures. We believe our proposed general device topology mapping algorithm will provide valuable information for architecture designers and assist them in designing a more DNN-friendly architecture.

关键词： Training Pipelines Topology Parallel processing Throughput Artificial neural networks network topology Deep learning system mapping neural network (NN) acceleration optimization parallelism partitioning

来源：评论

学校读者我要写书评

暂无评论

distributed Model-Free Adaptive Predictive Control for MIMO Multi-Agent Systems With Deception Attack

引用

IEEE TRANSACTIONS ON SIGNAL AND INFORMATION processing OVER networkS 2024年 10卷 32-47页

作者： Pan, Zhenzhen Chi, Ronghu Hou, Zhongsheng Qingdao Univ Dept Automat Acad Syst Sci & Control Qingdao 266071 Peoples R China Qingdao Univ Sci & Technol Inst Artificial Intelligence & Control Sch Automat & Elect Engn Qingdao 266061 Peoples R China

This work explores the challenging problems of nonlinear dynamics, nonaffine structures, heterogeneous properties, and deception attack together and proposes a novel distributed model-free adaptive predictive control (DMFAPC) for multiple-input-multiple-output (MIMO) multi-agent systems (MASs). A dynamic linearization method is introduced to address the nonlinear heterogeneous dynamics which is transformed as the unknown parameters in the obtained linear data model. A radial basis function neural network is designed to detect the deception attack and to estimate the polluted output that is further used in the controller design to compensate for the effect. Then, the DMFAPC is designed by defining a new expanded distributed output with a stochastic factor introduced. The bounded convergence is proved by using the contraction mapping method and the effectiveness of the proposed DMFAPC is verified by simulation examples.

关键词： Deception attack distributed model-free adaptive predictive control (DMFAPC) multi-agent systems (MASs) neural networks (NNs)

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：