期刊文献+
任意字段
题名或关键词
题名
关键词
摘要
作者
第一作者
机构
刊名
分类号
参考文献
作者简介
基金资助
栏目信息
共找到118,682篇文章
< 1 2 … 250 >
每页显示 20 50 100
已选0条
引用分析
参考文献
引证文献
 统计分析
检索结果
已选文献
显示方式：
Voxel event graph neural network for event-based human gait recognition 认领 引用
1
作者 Guangyuan MA Xiaolin GONG +2 位作者 Jiangtao XU Jiandong GAO Zhaoxuan GUO 《Optoelectronics Letters》 EI 2026年第3期167-173,共7页
To improve the accuracy of event-based gait recognition,a method called voxel event graph neural network(VEGNN)is proposed.This method voxelizes the event stream and selects representative voxels as vertices of the gr... To improve the accuracy of event-based gait recognition,a method called voxel event graph neural network(VEGNN)is proposed.This method voxelizes the event stream and selects representative voxels as vertices of the graph.Then the edges are constructed based on spatio-temporal distance and temporal order constraints so that the event stream is constructed as a graph structure.Finally,a lightweight feature extraction network based on graph neural networks(GNNs)is used to efficiently capture spatio-temporal information and motion cues from the event graph.To evaluate our method,an event-based gait recognition dataset called Celex-Gait is created.Experimental results show that VEGNN achieves a gait recognition accuracy of 95.8%and 93.9%on the Celex-Gait dataset and the DVS128-Gait-Day dataset,respectively.Furthermore,compared to the state-of-the-art methods,VEGNN reduces the number of model parameters by 30%.This indicates that VEGNN achieves higher recognition accuracy with lower model complexity. 展开更多
关键词 selects representative voxels voxel event graph neural network vegnn event stream lightweight feature extraction network voxel event graph neural network graph structurefinallya graph neural networks gnns gait recognitiona
暂未订购 下载PDF
Tensor Low-Rank Orthogonal Compression for Convolutional Neural Networks 认领 引用
2
作者 Yaping He Xin Luo 《IEEE/CAA Journal of Automatica Sinica》 SCIE EI CSCD 2026年第1期227-229,共3页
Dear Editor,The letter proposes a tensor low-rank orthogonal compression(TLOC)model for a convolutional neural network(CNN),which facilitates its efficient and highly-accurate low-rank representation.Model compression... Dear Editor,The letter proposes a tensor low-rank orthogonal compression(TLOC)model for a convolutional neural network(CNN),which facilitates its efficient and highly-accurate low-rank representation.Model compression is crucial for deploying deep neural network(DNN)models on resource-constrained embedded devices. 展开更多
关键词 model compression convolutional neural network cnn which tensor low rank orthogonal compression deep neural network dnn models embedded devices convolutional neural networks
暂未订购 下载PDF
Recognition and classification of microseismic signals based on Bayesian-optimized CNN-LSTM neural network 认领 引用
3
作者 WU Yang LIU Jian-feng +4 位作者 WANG Chun-ping LIU Jun-jie JI Zhen-xing TIAN Cheng-yu XUE Fu-jun 《Journal of Central South University》 SCIE EI CAS CSCD 2026年第6期2762-2787,共26页
Microseismic monitoring and signal recognition constitute critical technologies for accurately assessing rockburst risks and ensuring the safe construction of underground rock engineering.This study developed a"S... Microseismic monitoring and signal recognition constitute critical technologies for accurately assessing rockburst risks and ensuring the safe construction of underground rock engineering.This study developed a"Surface+Underground"microseismic intelligent monitoring system to evaluate dynamic disaster risks during construction at the Beishan High-level Radioactive Waste Geological Disposal Laboratory in China.The datasets of four typical one-dimensional time-domain microseismic signals,including rock fracture,blasting,TBM tunneling,and drilling,were constructed,and the BO-CNN-LSTM model is developed to identify and classify these signals..Based on the classification results,the typical time-frequency domain characteristics of the four types of signals are analyzed.The classification results of BO-CNN-LSTM,CNN,LSTM and CNN-LSTM models show that the recognition accuracy of the four models is 98.0%,85.7%,66.7%and 91.0%,respectively.Among all types,the four models demonstrate the highest effectiveness in identifying rock fracture and borehole signals.The findings further confirm that the BO-CNN-LSTM model efficiently recognizes and extracts microseismic signal features,demonstrating superior performance and stability in the classification task.Finally,the study suggests several future research directions,particularly in the areas of raw signal denoising,and automation of feature extraction. 展开更多
关键词 deep ground engineering microseismic monitoring signal classification neural network Bayesian optimization convolutional neural network(CNN) long short-term memory network(LSTM) BO-CNN-LSTM model
暂未订购 下载PDF
Personalized Differential Privacy Graph Neural Network 认领 引用
4
作者 Yanli Yuan Dian Lei +3 位作者 Chuan Zhang Zehui Xiong Chunhai Li Liehuang Zhu 《IEEE/CAA Journal of Automatica Sinica》 SCIE EI CSCD 2026年第2期498-500,共3页
Dear Editor,This letter addresses the critical challenge of preserving privacy in graph learning without compromising on data utility.Differential privacy(DP)is emerging as an effective method for privacy-preserving g... Dear Editor,This letter addresses the critical challenge of preserving privacy in graph learning without compromising on data utility.Differential privacy(DP)is emerging as an effective method for privacy-preserving graph learning.However,its application often diminishes data utility,especially for nodes with fewer neighbors in graph neural networks(GNNs). 展开更多
关键词 graph neural networks gnns personalized differential privacy graph learning privacy preservation data utility preserving privacy graph neural network
暂未订购 下载PDF
Structural Optimization of a Multi-Story Frame Structure Based on a Pre-Trained Physics-Informed Neural Network(PINN)Surrogate Model 认领 引用
5
作者 An Xu Zhixiong Liu +4 位作者 Hua Rong Liang Han Wei Shi Jun Huang Jiyang Fu 《Computer Modeling in Engineering & Sciences》 SCIE EI 2026年第4期279-313,共35页
In structural optimization,data-driven surrogate models are often explored as alternatives to finite element analysis to reduce computational cost.However,conventional neural networks usually fail to capture key struc... In structural optimization,data-driven surrogate models are often explored as alternatives to finite element analysis to reduce computational cost.However,conventional neural networks usually fail to capture key structural characteristics and are limited to predicting global responses(e.g.,top displacement),but usually fail to achieve accurate internal force predictions with conventional training data volumes.As a result,most existing studies involving surrogate models did not concern internal force constraints.To address this issue,this study proposes a structural optimization framework based on a pre-trained Physics-Informed Neural Network(PINN)surrogate model.By embedding static equilibrium equation into the loss function,the model achieves higher predictive accuracy,particularly for internal forces,while pre-training accelerates convergence and enhances stability.Combined with an improved multi-swarm particle swarm optimization(MPSO)algorithm,the framework enables efficient optimization of multi-story frame structures under internal force and multiple other constraints.The application to a six-story frame structure validates its effectiveness:compared with a DNN-based model,the PINN-based model improves the coefficient of determination for internal force prediction from 0.8874 to 0.9937.These results demonstrate that the proposed method offers a promising approach for efficient optimization of multi-story frame structures. 展开更多
关键词 Structural optimization intelligent algorithms physics-informed neural networks(PINN) neural networks
暂未订购 下载PDF
Experimental data-driven deep neural network modeling and prediction for the streaks of turbulent separated shear flow 认领 引用 被引量：1
6
作者 Xingyu Ma Jiateng Pan +1 位作者 Yihong Liu Nan Jiang 《Acta Mechanica Sinica》 SCIE EI CAS CSCD 2026年第7期52-58,共7页
In this paper,we report a recent experimental study of streak structures in the turbulent separated shear flow by datadriven deep neural network.By applying spanwise-aligned tetrahedron vortex generators upstream of a... In this paper,we report a recent experimental study of streak structures in the turbulent separated shear flow by datadriven deep neural network.By applying spanwise-aligned tetrahedron vortex generators upstream of a plane backward-facing step,spanwise-aligned high-and low-speed streaks were generated within the separated shear layer behind the step.The velocity profiles of the shear flow were measured by single-probe hot-wire anemometer in both the streamwise-vertical and the streamwise-spanwise planes in the wind tunnel.Deep neural network models are trained and verified based on the experimental datasets.The input parameter sets include the vortex generator height,spanwise spacing,and the spatial coordinates within the measurement domain,while the output parameter sets are mean and root-mean-square velocities of the shear flow.Mean squared errors between the model-predicted and experimentally measured data are used for quality evaluation of different deep neural network model designs,among which the minimum error of the optimal design descends less than 1%.For other vortex generator parameters,which are not measured in the wind tunnel or used in the training,the model prediction provides reasonable mean velocity contours with streak structures.Thus,we find that the experimental data-driven modeling approach shows reliable robustness for nonlinear fitting of complex datasets as well as considerable generalization for turbulent coherent structures. 展开更多
关键词 Streak Experimental data-driven Deep neural network Vortex generator Backward-facing step
暂未订购 下载PDF
Physics-informed neural networks for hidden boundary detection and flow field reconstruction 认领 引用 被引量：1
7
作者 Yongzheng Zhu Weizheng Chen +1 位作者 Jian Deng Xin Bian 《Acta Mechanica Sinica》 SCIE EI CAS CSCD 2026年第7期131-149,共19页
Simultaneously detecting hidden solid boundaries and reconstructing flow fields from sparse observations poses a significant inverse challenge in fluid mechanics.This study presents a physics-informed neural network f... Simultaneously detecting hidden solid boundaries and reconstructing flow fields from sparse observations poses a significant inverse challenge in fluid mechanics.This study presents a physics-informed neural network framework designed to infer the presence,shape,and motion of static or moving solid boundaries within a flow field.By integrating a body fraction parameter into the governing equations,the model enforces no-slipo-penetration boundary conditions in solid regions while preserving conservation laws of fluid dynamics.Using partial flow field data,the method simultaneously reconstructs the unknown flow field and infers the body fraction distribution,thereby revealing solid boundaries.The framework is validated across diverse scenarios,including incompressible Navier-Stokes and compressible Euler flows,such as steady flow past a fixed cylinder,an inline oscillating cylinder,and subsonic flow over an airfoil.The results demonstrate accurate detection of hidden boundaries,reconstruction of missing flow data,and estimation of trajectories and velocities of a moving body.Further analysis examines the effects of data sparsity,velocity-only measurements,and noise on inference accuracy.The proposed method exhibits robustness and versatility,highlighting its potential for applications when only limited experimental or numerical data are available. 展开更多
关键词 Hidden moving boundary Flow reconstruction Inverse problem Physics-informed neural networks
暂未订购 下载PDF
Artificial Neural Network-Based Prediction and Validation of Drill Flank Wear in GFRP Machining for Sustainable and Smart Manufacturing 认领 引用 被引量：1
8
作者 Sathish Rao Udupi Gururaj Bolar +1 位作者 Manjunath Shettar Ashwini Bhat 《Computers, Materials & Continua》 SCIE EI 2026年第6期903-918,共16页
Glass fiber-reinforced polymer composites(GFRPCs)are extensively utilized in the aerospace,automotive,and structural sectors;nevertheless,their heterogeneous and abrasive characteristics result in rapid tool wear duri... Glass fiber-reinforced polymer composites(GFRPCs)are extensively utilized in the aerospace,automotive,and structural sectors;nevertheless,their heterogeneous and abrasive characteristics result in rapid tool wear during drilling.Drill flank wear among various wear mechanisms notably influences hole quality and dimensional accuracy.This research investigates the impact of spindle speed,feed rate,and drill diameter on flank wear during dry drilling of GFRPC laminates with high-speed steel(HSS)twist drills.A full-factorial design with 81 experiments is used to create a comprehensive dataset.ANOVA indicates that spindle speed is the dominant factor affecting wear changes,accounting for 74.43%,followed by feed rate(15.80%)and drill diameter(6.16%).A linear regression model demonstrates reasonable statistical sufficiency(R2=0.964),but it falls short in reflecting nonlinear interactions.Hence,an artificial neural network(ANN)model is developed to improve prediction.The multilayer feed-forward ANN with a 3-10-6-1 architecture,trained using the Levenberg-Marquardt optimization algorithm,achieves excellent predictive accuracy,with high correlation and low root-mean-square error.Model validation was achieved through independent confirmation experiments,yielding a mean absolute percentage error of only 2.27%,with all predictions falling within the permissible wear range.The findings indicate that ANN-based modeling provides a reliable framework for capturing the complex nonlinear relationships governing tool wear in GFRPC drilling and serves as a viable soft sensor for tool condition monitoring,process optimization,and sustainable,data-driven manufacturing. 展开更多
关键词 Glass fiber-reinforced polymer composites machining and drilling drill flank wear prediction artificial neural networks(ANN) soft computational modeling
暂未订购 下载PDF
Learning Laws for Deep Convolutional Neural Networks With Guaranteed Convergence 认领 引用 被引量：1
9
作者 Sitan Li Chien Chern Cheah 《IEEE/CAA Journal of Automatica Sinica》 SCIE EI CSCD 2026年第1期170-185,共16页
Convolutional neural networks(CNNs)have shown remarkable success across numerous tasks such as image classification,yet the theoretical understanding of their convergence remains underdeveloped compared to their empir... Convolutional neural networks(CNNs)have shown remarkable success across numerous tasks such as image classification,yet the theoretical understanding of their convergence remains underdeveloped compared to their empirical achievements.In this paper,the first filter learning framework with convergence-guaranteed learning laws for end-to-end learning of deep CNNs is proposed.Novel update laws with convergence analysis are formulated based on the mathematical representation of each layer in convolutional neural networks.The proposed learning laws enable concurrent updates of weights across all layers of the deep convolutional neural network and the analysis shows that the training errors converge to certain bounds which are dependent on the approximation errors.Case studies are conducted on benchmark datasets and the results show that the proposed concurrent filter learning framework guarantees the convergence and offers more consistent and reliable results during training with a trade-off in performance compared to stochastic gradient descent methods.This framework represents a significant step towards enhancing the reliability and effectiveness of deep convolutional neural network by developing a theoretical analysis which allows practical implementation of the learning laws with automatic tuning of the learning rate to guarantee the convergence during training. 展开更多
关键词 Convergence convolution neural networks(CNNs) end-to-end learning online learning
暂未订购 下载PDF
Neural network control for mitigating actuator delay in ATR engines using predictive compensation and stability reward 认领 引用
10
作者 Weidong CAI Wei ZHAO +3 位作者 Xiaorong XIANG Sanqun REN Xuesen YANG Qingjun ZHAO 《Chinese Journal of Aeronautics》 SCIE EI CAS CSCD 2026年第2期301-327,共27页
The flight envelope of Air Turbo Rocket(ATR)engines is broader compared to conventional aero-engines,and designing a full-envelope controller using traditional methods poses significant challenges due to a burdensome ... The flight envelope of Air Turbo Rocket(ATR)engines is broader compared to conventional aero-engines,and designing a full-envelope controller using traditional methods poses significant challenges due to a burdensome design process.To address this issue,this paper proposes a self-learning neural network controller design method based on Reinforcement Learning(RL).Additionally,a method for predictive compensation and stability rewards is proposed to reduce the system oscillation caused by actuator delay.This approach simplifies the actuator to a firstorder inertial element exhibiting pure delay.A simulation environment for the ATR engineactuator system is first established.Based on this environment,a self-learning neural network controller using a predictive compensator and the Proximal Policy Optimization(PPO)algorithm is then developed.Furthermore,the temporal difference signals from the controller output are integrated into the reward function to enhance system stability.The proposed method is validated through numerical simulations and semi-physical experiments.The numerical simulation results demonstrate that the proposed method increases the system's tolerance to delays from 20 ms to 400 ms.Under an actuator delay of 400 ms,the average steady-state error remains less than0.1%,the overshoot is limited to 1%,and the settling time does not exceed 3 s.Moreover,compared to the traditional method,the proposed method exhibits higher adaptability to model errors and variations in flight conditions.In the conducted semi-physical simulation experiments,the proposed method achieves stable control of a real electric pump. 展开更多
关键词 Actuator delay ATR engines Neural network control Reinforcement learning Semi-physical simulation
暂未订购 下载PDF
A neural network response surface updating method for complex dynamic models based on modal matching reconstruction strategy 认领 引用
11
作者 Xiang LI Zhinong JIANG +2 位作者 Huajin SHAO Hao WANG Yanfei ZUO 《Chinese Journal of Aeronautics》 SCIE EI CAS CSCD 2026年第5期626-642,共17页
An integrated dynamic model updating method is proposed to address the challenges of mode jumping and mode degeneracy for complex dynamic models.A reduced order proxy model of Neural Network Response Surface(NNRS)was ... An integrated dynamic model updating method is proposed to address the challenges of mode jumping and mode degeneracy for complex dynamic models.A reduced order proxy model of Neural Network Response Surface(NNRS)was constructed by Modal Matching Reconstruction Strategy(MMRS)and an Improved Vectorial Surrogate Model(IVSM).Among them,the analytical modes are correctly matched with the experimental modes by MMRS,and the order of the mode matching is determined by calculating the Modal Assurance Criterion(MAC),addressing the dynamic changes of the mode matching order during the construction of the NNRS.The fitted NNRS model results are vectorized by IVSM,enabling the rapid extraction of required input and output parameters under multi-parameter conditions,thereby improving efficiency.The model parameters are updated using a multi-objective genetic algorithm,which achieves the simultaneous updating of natural frequency and mode shape.To validate the accuracy and efficiency,an intermediate casing of a gas turbine was updated using the proposed method.With high efficiency,the mean absolute error of natural frequency for the matched order decreased from 24.46%to 3.89%,while the corresponding average MAC value increased from 0.654 to 0.752. 展开更多
关键词 Complex structures Dynamic model updating Modal analysis Modal Assurance Criterion Neural network Response surface
暂未订购 下载PDF
Joint physics and model-guided pre-stack seismic inversion using double dual neural network 认领 引用
12
作者 Jian Zhang Hui Sun +2 位作者 Xing-Guo Huang Li Han Yan-Song Li 《Petroleum Science》 SCIE EI CAS CSCD 2026年第1期157-177,共21页
Pre-stack seismic inversion is used to calculate elastic parameters,including P-wave and S-wave velocities,as well as densities.These parameters play an integral role in the characterization of reservoirs,thereby enha... Pre-stack seismic inversion is used to calculate elastic parameters,including P-wave and S-wave velocities,as well as densities.These parameters play an integral role in the characterization of reservoirs,thereby enhancing the exploration and production process.Deep learning-based seismic inversion does not need a known physical system and can give satisfactory results with sufficient training data.The acquisition of such datasets for seismic inversion poses a significant challenge due to the exorbitant costs associated with drilling activities.Integrating domain knowledge,physical systems,and well log data into a deep learning-based seismic inversion framework is crucial for improving its efficiency and effectiveness.Nevertheless,existing data-driven approaches do not adequately exploit such information,thereby constraining their overall performance and applicability.Therefore,we develop a double dual neural network structure built upon the closed-loop neural network framework,which incorporates both physics and model information to mitigate the dependency on extensive labeled datasets.The information from the different domains is linked through a loss function,where one dual network is responsible for constraining the inversion results using physics information to ensure the physics consistency of the predictions,and the other dual network is responsible for constraining the inversion results using a priori model information to enhance the reliability of the predictions.The method makes full use of well-log data for network training when wells are available,as well as providing unsupervised learning and inversion under well-free conditions.The integration of qualitative and quantitative analyses proves instrumental in demonstrating the effectiveness of the proposed methodology through the use of synthetic and field pre-stack examples. 展开更多
关键词 Pre-stack seismic inversion Double dual neural network Physics guided Model guided
暂未订购 下载PDF
Meson Properties and Symmetry Emergence Based on the Deep Neural Network 认领 引用
13
作者 Xin Tong Wei Feng +3 位作者 Weiwei Xu Chao-Hsi Chang Guo-Li Wang Qiang Li 《Chinese Physics Letters》 SCIE EI CAS CSCD 2026年第2期32-50,共19页
As a key property of hadrons,the total width is quite difficult to obtain in theory due to the extreme complexity of the strong and electroweak interactions.In this work,a deep neural network model with the Transforme... As a key property of hadrons,the total width is quite difficult to obtain in theory due to the extreme complexity of the strong and electroweak interactions.In this work,a deep neural network model with the Transformer architecture is built to precisely predict meson widths in the range of 10-14-625 Me V based on meson quantum numbers and masses.The relative errors of the predictions are 0.12%,2.0%,and 0.54% in the training set,the test set,and all the data,respectively.We present the predicted meson width spectra for the currently discovered states and some theoretically predicted ones.The model is also used as a probe to study the quantum numbers and inner structures for some undetermined states,including the exotic states.Notably,this data-driven model is found to spontaneously exhibit good charge conjugation symmetry and approximate isospin symmetry consistent with physical principles.The results indicate that the deep neural network can serve as an independent complementary research paradigm to describe and explore the hadron structures and the complicated interactions in particle physics alongside traditional experimental measurements,theoretical calculations,and lattice simulations. 展开更多
关键词 transformer architecture charge conjugation symmetry deep neural network model symmetry emergence deep neural network meson widths strong electroweak interactionsin meson properties
暂未订购 下载PDF
Physics-informed neural network model for predicting projectile mass erosion integrating cutting and thermal melting mechanisms 认领 引用
14
作者 Hengwei Xu Yonggang Lu +3 位作者 Lingmei Guo Junrun Li Xiaowei Feng Zhengcao Lu 《Defence Technology（防务技术）》 SCIE EI CAS CSCD 2026年第5期310-324,共15页
High-velocity penetration of projectiles into concrete induces intense thermo mechanical interactions that lead to substantial projectile mass erosion,thereby compromising structural integrity and ballistic stability.... High-velocity penetration of projectiles into concrete induces intense thermo mechanical interactions that lead to substantial projectile mass erosion,thereby compromising structural integrity and ballistic stability.Accurate mass erosion prediction is therefore critical for warhead lethality assessment.To address this challenge,we develop a physics-informed neural network(PINN)model for predicting projectile mass erosion at velocities ranging from 345 m/s to 1852 m/s.The model integrates physicsbased constraints from cutting and thermal melting mechanisms through a composite loss function.Data augmentation and dimensionality reduction techniques are applied to improve model generalization across diverse impact conditions.Training and validation are conducted on an augmented dataset comprising 283 samples,derived from 10 published studies.The model's generalization capability is rigorously evaluated on an independent set of 13 new penetration experiments with ogivenosed projectiles made of three distinct materials.Compared to a purely data-driven neural network(NN)model,the PINN model reduces the average relative error on the validation set from 16.3%to 14.5%and improves the proportion of predictions within a 20%relative error margin from 67.3%to 71.2%.On an independent test set,the PINN model demonstrates superior accuracy over both the data-driven neural network model and conventional theoretical models,achieving an average relative error of11.7%,with 38.5%and 84.6%of the predictions falling within 10%and 20%relative error margins,respectively.By enabling accurate mass erosion prediction under high-velocity conditions,the proposed methodology supports weight-optimized penetrator design and directly contributes to terminal effectiveness improvement. 展开更多
关键词 High-velocity penetration Projectile mass erosion Machine learning Physics-informed neural network
暂未订购 下载PDF
A heterogeneous cyclic Hopfield neural network without self-connections 认领 引用
15
作者 Yunzhen ZHANG Chunlong ZHOU +2 位作者 Han BAO Guangzhe ZHAO Bocheng BAO 《Journal of Zhejiang University-SCIENCE A》 SCIE EI CAS CSCD 2026年第3期306-316,共11页
We propose a three-neuron heterogeneous cyclic Hopfield neural network(het-CHNN)utilizing three different activation functions:the hyperbolic tangent,sine,and cosine functions.The network’s globally uniformly ultimat... We propose a three-neuron heterogeneous cyclic Hopfield neural network(het-CHNN)utilizing three different activation functions:the hyperbolic tangent,sine,and cosine functions.The network’s globally uniformly ultimate boundedness is proved theoretically,and its chaotic dynamics are explored through numerical simulations and analog experiments.The numerical results demonstrate that the het-CHNN displays chaotic dynamics and multi-scroll chaotic attractors.Subsequently,the het-CHNN is implemented in an analog circuit,and hardware experiments are performed to verify the previous numerical results.Notably,the het-CHNN successfully resolves the issue of the absence of chaos in a three-neuron CHNN and currently appears to be the simplest three-neuron Hopfield neural network(HNN)that can generate chaos. 展开更多
关键词 Activation function Analog circuit Chaos Cyclic Hopfield neural network(CHNN) Multi-scroll chaotic attractor
暂未订购 下载PDF
A quantitative interpretation method of gas kick driven by physicsinformed neural network 认领 引用
16
作者 Hong-Wei Yang Biao Wang +5 位作者 Jun Li Geng Zhang Gong-Hui Liu Jia-Hao Zhan Zhen-Yu Long Chao Wang 《Petroleum Science》 SCIE EI CAS CSCD 2026年第5期2599-2621,共23页
To address the challenges associated with predicting wellbore fluid flow behavior and gas kick rates in deep,complex formations following gas kick events,this study develops a quantitative interpretation method of gas... To address the challenges associated with predicting wellbore fluid flow behavior and gas kick rates in deep,complex formations following gas kick events,this study develops a quantitative interpretation method of gas kick driven by physics-informed neural network(PINN).The proposed method integrates a physical model of gas—liquid two-phase flow in the wellbore into the neural network by formulating it as a loss function,leveraging annulus temperature and pressure data obtained from downhole dual measurement tools.The feasibility and effectiveness of this method are evaluated through comparative analysis.The result indicates that duringgas kick occurrences,this method achieves mean relative errors of 8.49%and 9.07%for the predicted gas volume fraction and apparent gas phase velocity between the dual measurement points,respectively,and 3.76%for the bottomhole gas kick rate,without the need for mesh discretization or predefined initial conditions,demonstrating strong applicability in field scenarios.Compared to the unscented Kalman filter(UKF)and genetic algorithm(GA),this method exhibits higher prediction accuracy and stability due to its global optimization capability,overcoming the divergence issues encountered by UKF and GA during point-wise recursive predictions under noisy pressure data conditions.Integrating this method with downhole dual measurement tools can provide valuable guidance for blowout risk assessment,well-control method selection,and well-killing parameter design after a gas kick. 展开更多
关键词 Gas kick Quantitative interpretation Physics-informed neural network Downhole dual measurement tools Partial differential equation
暂未订购 下载PDF
Model predictive control based on LSTM neural network for maglev vehicle’suspension system 认领 引用
17
作者 Mengjuan Liu Han Wu +3 位作者 Xin Liang Jiali Liu Xiaohui Zeng Kaixuan Hu 《Acta Mechanica Sinica》 SCIE EI CAS CSCD 2026年第5期504-517,共14页
To improve the suspension performance of high-speed maglev vehicles under complex external disturbance,a composite model predictive control(MPC)algorithm based on a neural network is proposed.Firstly,the nonlinear dyn... To improve the suspension performance of high-speed maglev vehicles under complex external disturbance,a composite model predictive control(MPC)algorithm based on a neural network is proposed.Firstly,the nonlinear dynamic response prediction model is constructed utilizing the long short-term memory(LSTM)neural network,and this model is trained by machine learning.Subsequently,a rolling optimization controller of the MPC algorithm is designed according to the vehicle suspension system’s prediction model and the suspension target.To compensate for the error of the prediction model resulting from changes in the control algorithm,a composite MPC algorithm is devised by combining both the proportionalintegral-derivative(PID)algorithm and the MPC algorithm.This composite approach enables the suspension system to switch the selection of control algorithms in the suspension system according to the prediction error.Finally,the effectiveness of the composite MPC algorithm is verified by simulation and experiment.The results show that the prediction model based on the LSTM neural network can effectively predict the future dynamic response of the vehicle.Moreover,the proposed MPC algorithm can effectively suppress the suspension gap fluctuation in the high-speed maglev vehicle,thereby fostering improved stability in the suspension system. 展开更多
关键词 Maglev vehicle Suspension system Model predictive control Long short-term memory neural network
暂未订购 下载PDF
PASSIVITY-BASED SYNCHRONIZATION OF MARKOVIAN JUMP INERTIAL NEURAL NETWORKS VIA ADAPTIVE EVENT-DRIVEN PROTOCOL AND ITS APPLICATION TO IMAGE ENCRYPTION 认领 引用
18
作者 A.CHANDRASEKAR Quanxin ZHU +2 位作者 T.RADHIKA M.KAMALI A.R.SUBHASHRI 《Acta Mathematica Scientia》 SCIE CSCD 2026年第3期1555-1580,共26页
This paper concentrates on the study of passivity-based synchronization of inertial neural networks including Markov jump parameters.The second-order differential equations are converted into first-order differential ... This paper concentrates on the study of passivity-based synchronization of inertial neural networks including Markov jump parameters.The second-order differential equations are converted into first-order differential equations using the variable transformation method.To make effective use of network bandwidth resources and to optimize the Markov jump inertial neural networks(MJINNs)performance,an adaptive event-driven protocol controller is studied.To achieve synchronization,an appropriate Lyapunov-Krasovskii functional(LKF)is constructed,which includes double integral terms that capture the information of time-varying delay terms.Some sufficient conditions are obtained in terms of linear matrix inequalities(LMIs)using Reciprocal convex combination lemma(RCCL).Then,a numerical simulation and an application of image encryption are carried out to illustrate the effectiveness of the proposed method. 展开更多
关键词 inertial neural networks adaptive event-driven protocol Markovian jump process synchronization
暂未订购 下载PDF
Residual resampling-based physics-informed neural network for neutron diffusion equations 认领 引用
19
作者 Heng Zhang Yun-Ling He +3 位作者 Dong Liu Qin Hang He-Min Yao Di Xiang 《Nuclear Science and Techniques》 SCIE EI CAS CSCD 2026年第2期16-41,共26页
The neutron diffusion equation plays a pivotal role in nuclear reactor analysis.Nevertheless,employing the physics-informed neural network(PINN)method for its solution entails certain limitations.Conventional PINN app... The neutron diffusion equation plays a pivotal role in nuclear reactor analysis.Nevertheless,employing the physics-informed neural network(PINN)method for its solution entails certain limitations.Conventional PINN approaches generally utilize a fully connected network(FCN)architecture that is susceptible to overfitting,training instability,and gradient vanishing as the network depth increases.These challenges result in accuracy bottlenecks in the solution.In response to these issues,the residual-based resample physics-informed neural network(R2-PINN)is proposed.It is an improved PINN architecture that replaces the FCN with a convolutional neural network with a shortcut(S-CNN).It incorporates skip connections to facilitate gradient propagation between network layers.Additionally,the incorporation of the residual adaptive resampling(RAR)mechanism dynamically increases the number of sampling points.This,in turn,enhances the spatial representation capabilities and overall predictive accuracy of the model.The experimental results illustrate that our approach significantly improves the convergence capability of the model and achieves high-precision predictions of the physical fields.Compared with conventional FCN-based PINN methods,R 2-PINN effectively overcomes the limitations inherent in current methods.Thus,it provides more accurate and robust solutions for neutron diffusion equations. 展开更多
关键词 Neutron diffusion equation Physics-informed neural network CNN with shortcut Residual adaptive resampling
暂未订购 下载PDF
Perceiving the Battery Multi-Electrochemical States in Real-time Based on Model-Informed Neural Network 认领 引用
20
作者 Ying Zhang Qinghao Zhang +3 位作者 Bin Duan Pingwei Gu Changlong Li Chenghui Zhang 《IEEE/CAA Journal of Automatica Sinica》 SCIE EI CSCD 2026年第5期1041-1053,共13页
Accurate estimation of electrochemical states serves as a pathway to observe internal battery behaviors,effectively bridging the gap between micro mechanism and macro performance and enabling more precise control in a... Accurate estimation of electrochemical states serves as a pathway to observe internal battery behaviors,effectively bridging the gap between micro mechanism and macro performance and enabling more precise control in an advanced battery management system.Yet conventional pseudo-two-dimensional(P2D)physics methods suffer from high computational complexity and limit their online application.Thus,we develop a modelinformed neural network(MINN)framework that synergistically combines deep learning with a physics-based model to accurately monitor the battery electrochemical state(such as lithium-ion concentration,plating potential).Firstly,the MINN model is constructed with the innovative loss term containing experimentally measurable parameters and governing physical laws.Secondly,a composite framework based on a convolutional neural network(CNN)architecture is integrated to automatically extract features and enforce spatial boundary conditions,which significantly reduces the number of boundary loss terms that need to be solved and alleviates the complexity of the training process.After training,the MINN model can achieve an accurate estimation of internal states and even their spatiotemporal distributions that cannot be directly measured based on limited observable data and physical laws.At last,by incorporating dynamic current input,the well-trained basic model exhibits strong robustness and can be directly transferred to other cycling protocols with high accuracy,requiring no further retraining.MINN is a novel and promising framework to realize online and accurate micro electrochemical states monitoring,achieving at least 776 times speedup compared with the P2D model.As an innovative artificial intelligence assisted modeling for electrochemical systems,this framework enables root-cause analysis of battery behavior and failure modes,while empowering the management system with more reliable and trustworthy decision-making capabilities. 展开更多
关键词 Electrochemical state estimation lithium-ion battery model-informed neural network(MINN) real-time application
暂未订购 下载PDF
已选0条
引用分析
参考文献
引证文献
 统计分析
检索结果
已选文献
上一页 1 2 … 250 下一页 到第 页
在线咨询 使用帮助 返回顶部 意见反馈