The deep spatiotemporal network with dual-flow fusion for video-oriented facial expression recognition

The video-oriented facial expression recognition has always been an important issue in emotion perception.At present,the key challenge in most existing methods is how to effectively extract robust features to characterize facial appearance and geometry changes caused by facial motions.On this basis,the video in this paper is divided into multiple segments,each of which is simultaneously described by optical flow and facial landmark trajectory.To deeply delve the emotional information of these two representations,we propose a Deep Spatiotemporal Network with Dual-flow Fusion(defined as DSN-DF),which highlights the region and strength of expressions by spatiotemporal appearance features and the speed of change by spatiotemporal geometry features.Finally,experiments are implemented on CKþand MMI datasets to demonstrate the superiority of the proposed method.

Adaptation analysis and fusion correction method of CMIP6 precipitation simulation data on the Qinghai-Tibetan Plateau

In order to obtain more accurate precipitation data and better simulate the precipitation on the Tibetan Plateau,the simulation capability of 14 Coupled Model Intercomparison Project Phase 6(CMIP6)models of historical precipitation(1982-2014)on the Qinghai-Tibetan Plateau was evaluated in this study.Results indicate that all models exhibit an overestimation of precipitation through the analysis of the Taylor index,temporal and spatial statistical parameters.To correct the overestimation,a fusion correction method combining the Backpropagation Neural Network Correction(BP)and Quantum Mapping(QM)correction,named BQ method,was proposed.With this method,the historical precipitation of each model was corrected in space and time,respectively.The correction results were then analyzed in time,space,and analysis of variance(ANOVA)with those corrected by the BP and QM methods,respectively.Finally,the fusion correction method results for each model were compared with the Climatic Research Unit(CRU)data for significance analysis to obtain the trends of precipitation increase and decrease for each model.The results show that the IPSL-CM6A-LR model is relatively good in simulating historical precipitation on the Qinghai-Tibetan Plateau(R=0.7,RSME=0.15)among the uncorrected data.In terms of time,the total precipitation corrected by the fusion method has the same interannual trend and the closest precipitation values to the CRU data;In terms of space,the annual average precipitation corrected by the fusion method has the smallest difference with the CRU data,and the total historical annual average precipitation is not significantly different from the CRU data,which is better than BP and QM.Therefore,the correction effect of the fusion method on the historical precipitation of each model is better than that of the QM and BP methods.The precipitation in the central and northeastern parts of the plateau shows a significant increasing trend.The correlation coefficients between monthly precipitation and site-detected precipitation for all models after BQ correction exceed 0.8.

Multi-scale persistent spatiotemporal transformer for long-term urban traffic flow prediction

Long-term urban traffic flow prediction is an important task in the field of intelligent transportation,as it can help optimize traffic management and improve travel efficiency.To improve prediction accuracy,a crucial issue is how to model spatiotemporal dependency in urban traffic data.In recent years,many studies have adopted spatiotemporal neural networks to extract key information from traffic data.However,most models ignore the semantic spatial similarity between long-distance areas when mining spatial dependency.They also ignore the impact of predicted time steps on the next unpredicted time step for making long-term predictions.Moreover,these models lack a comprehensive data embedding process to represent complex spatiotemporal dependency.This paper proposes a multi-scale persistent spatiotemporal transformer(MSPSTT)model to perform accurate long-term traffic flow prediction in cities.MSPSTT adopts an encoder-decoder structure and incorporates temporal,periodic,and spatial features to fully embed urban traffic data to address these issues.The model consists of a spatiotemporal encoder and a spatiotemporal decoder,which rely on temporal,geospatial,and semantic space multi-head attention modules to dynamically extract temporal,geospatial,and semantic characteristics.The spatiotemporal decoder combines the context information provided by the encoder,integrates the predicted time step information,and is iteratively updated to learn the correlation between different time steps in the broader time range to improve the model’s accuracy for long-term prediction.Experiments on four public transportation datasets demonstrate that MSPSTT outperforms the existing models by up to 9.5%on three common metrics.

Research on simulation of gun muzzle flow field empowered by artificial intelligence

Artificial intelligence technology is introduced into the simulation of muzzle flow field to improve its simulation efficiency in this paper.A data-physical fusion driven framework is proposed.First,the known flow field data is used to initialize the model parameters,so that the parameters to be trained are close to the optimal value.Then physical prior knowledge is introduced into the training process so that the prediction results not only meet the known flow field information but also meet the physical conservation laws.Through two examples,it is proved that the model under the fusion driven framework can solve the strongly nonlinear flow field problems,and has stronger generalization and expansion.The proposed model is used to solve a muzzle flow field,and the safety clearance behind the barrel side is divided.It is pointed out that the shape of the safety clearance under different launch speeds is roughly the same,and the pressure disturbance in the area within 9.2 m behind the muzzle section exceeds the safety threshold,which is a dangerous area.Comparison with the CFD results shows that the calculation efficiency of the proposed model is greatly improved under the condition of the same calculation accuracy.The proposed model can quickly and accurately simulate the muzzle flow field under various launch conditions.

Towards implementation of alloy-specific thermo-fluid modelling for laser powder-bed fusion of Mg alloys

Multi-physics thermo-fluid modeling has been extensively used as an approach to understand melt pool dynamics and defect formation as well as optimizing the process-related parameters of laser powder-bed fusion(L-PBF).However,its capabilities for being implemented as a reliable tool for material design,where minor changes in material-related parameters must be accurately captured,is still in question.In the present research,first,a thermo-fluid computational fluid dynamics(CFD)model is developed and validated against experimental data.Considering the predicted material properties of the pure Mg and commercial ZK60 and WE43 Mg alloys,parametric studies are done attempting to elucidate how the difference in some of the material properties,i.e.,saturated vapor pressure,viscosity,and solidification range,can influence the melt pool dynamics.It is found that a higher saturated vapor pressure,associated with the ZK60 alloy,leads to a deeper unstable keyhole,increasing the keyhole-induced porosity and evaporation mass loss.Higher viscosity and wider solidification range can increase the non-uniformity of temperature and velocity distribution on the keyhole walls,resulting in increased keyhole instability and formation of defects.Finally,the WE43 alloy showed the best behavior in terms of defect formation and evaporation mass loss,providing theoretical support to the extensive use of this alloy in L-PBF.In summary,this study suggests an approach to investigate the effect of materials-related parameters on L-PBF melting and solidification,which can be extremely helpful for future design of new alloys suitable for L-PBF.

A multi-source data fusion modeling method for debris flow prevention engineering

The Digital Elevation Model(DEM)data of debris flow prevention engineering are the boundary of a debris flow prevention simulation,which provides accurate and reliable DEM data and is a key consideration in debris flow prevention simulations.Thus,this paper proposes a multi-source data fusion method.First,we constructed 3D models of debris flow prevention using virtual reality technology according to the relevant specifications.The 3D spatial data generated by 3D modeling were converted into DEM data for debris flow prevention engineering.Then,the accuracy and applicability of the DEM data were verified by the error analysis testing and fusion testing of the debris flow prevention simulation.Finally,we propose the Levels of Detail algorithm based on the quadtree structure to realize the visualization of a large-scale disaster prevention scene.The test results reveal that the data fusion method controlled the error rate of the DEM data of the debris flow prevention engineering within an allowable range and generated 3D volume data(obj format)to compensate for the deficiency of the DEM data whereby the 3D internal entity space is not expressed.Additionally,the levels of detailed method can dispatch the data of a large-scale debris flow hazard scene in real time to ensure a realistic 3D visualization.In summary,the proposed methods can be applied to the planning of debris flow prevention engineering and to the simulation of the debris flow prevention process.

MHD Stability Analysis and Flow Controls of Liquid Metal Free Surface Film Flows as Fusion Reactor PFCs

Numerical and experimental investigation results on the magnetohydrodynamics（MHD） film flows along flat and curved bottom surfaces are summarized in this study. A simplified modeling has been developed to study the liquid metal MHD film state, which has been validated by the existing experimental results. Numerical results on how the inlet velocity（V）, the chute width（W） and the inlet film thickness（d0） affect the MHD film flow state are obtained. MHD stability analysis results are also provided in this study. The results show that strong magnetic fields make the stable V decrease several times compared to the case with no magnetic field,especially small radial magnetic fields（Bn） will have a significant impact on the MHD film flow state. Based on the above numerical and MHD stability analysis results flow control methods are proposed for flat and curved MHD film flows. For curved film flow we firstly proposed a new multi-layers MHD film flow system with a solid metal mesh to get the stable MHD film flows along the curved bottom surface. Experiments on flat and curved MHD film flows are also carried out and some firstly observed results are achieved.

Identification Method of Gas-Liquid Two-phase Flow Regime Based on Image Multi-feature Fusion and Support Vector Machine

流动政体的知识为确定是很重要的二阶段的流动系统的压力落下，稳定性和安全。基于图象多特征熔化和支持向量机器，在二阶段的流动识别流动政体的一个新方法被介绍。第一，煤气液体的二阶段的流动想象包括冒泡的流动，塞子流动，蛞蝓流动，成层的流动，起浪的流动，环形的流动和薄雾流动被数字高速度录像系统在水平试管捕获。图象时刻 invariants 和灰色的水平同现矩阵质地特征用处理技术的图象被提取。改进一个多重分类器系统的表演，不平的集合理论被用于减少无关紧要的因素。而且，支持向量机器被使用这些特徵向量作为流动政体样品，和流动政体减少尺寸训练聪明的鉴定被认识到。测试结果证明与不平的集合理论被减少的图象特征能最优地反映支持向量机器能快速并且精确地识别的七典型流动政体，并且成功的训练之间的差别在水平试管的煤气液体的二阶段的流动的七典型流动政体。图象多特征熔化方法提供了一个新方法识别煤气液体的二阶段的流动，并且比单个特征的完成了更高的鉴定能力。全面鉴定精确性是 100% ，并且处理时间的图象的估计是为联机流动政体鉴定的 8 ms。

A brief review：experimental investigation of zonal flows and geodesic acoustic modes in fusion plasmas

Zonal flows self-generated by turbulence play an important role in regulating turbulence,reducing transport level,and thus improve plasma confinement in fusion plasmas.The zonal flows and geodesic acoustic modes have been identified in various devices.The related issues,such as the poloidal and toroidal symmetries,coupling to turbulence,effects on turbulence and transport,nonlinear energy transfer between turbulence and zonal flows,dependence of the plasma parameters,roles in the confinement regime transitions etc are overviewed briefly in this paper.The interaction between zonal flows and magnetic islands is emphasized.

DuFNet:Dual Flow Network of Real-Time Semantic Segmentation for Unmanned Driving Application of Internet of Things

The application of unmanned driving in the Internet of Things is one of the concrete manifestations of the application of artificial intelligence technology.Image semantic segmentation can help the unmanned driving system by achieving road accessibility analysis.Semantic segmentation is also a challenging technology for image understanding and scene parsing.We focused on the challenging task of real-time semantic segmentation in this paper.In this paper,we proposed a novel fast architecture for real-time semantic segmentation named DuFNet.Starting from the existing work of Bilateral Segmentation Network(BiSeNet),DuFNet proposes a novel Semantic Information Flow(SIF)structure for context information and a novel Fringe Information Flow(FIF)structure for spatial information.We also proposed two kinds of SIF with cascaded and paralleled structures,respectively.The SIF encodes the input stage by stage in the ResNet18 backbone and provides context information for the feature fusionmodule.Features from previous stages usually contain rich low-level details but high-level semantics for later stages.Themultiple convolutions embed in Parallel SIF aggregate the corresponding features among different stages and generate a powerful global context representation with less computational cost.The FIF consists of a pooling layer and an upsampling operator followed by projection convolution layer.The concise component provides more spatial details for the network.Compared with BiSeNet,our work achieved faster speed and comparable performance with 72.34%mIoU accuracy and 78 FPS on Cityscapes Dataset based on the ResNet18 backbone.

Flow field fusion simulation method based on model features and its application in CRDM

The control rod drive mechanism(CRDM)is an essential part of the control and safety protection system of pressurized water reactors.Current CRDM simulations are mostly performed collectively using a single method,ignoring the influence of multiple motion units and the differences in various features among them,which strongly affect the efficiency and accuracy of the simulations.In this study,we constructed a flow field fusion simulation method based on model features by combining key motion unit analysis and various simulation methods and then applied the method to the CRDM simulation process.CRDM performs motion unit decomposition through the structural hierarchy of function-movement-action method,and the key meta-actions are identified as the nodes in the flow field simulation.We established a fused feature-based multimethod simulation process and processed the simulation methods and data according to the features of the fluid domain space and the structural complexity to obtain the fusion simulation results.Compared to traditional simulation methods and real measurements,the simulation method provides advantages in terms of simulation efficiency and accuracy.

Investigation of the Laser Powder Bed Fusion Process of Ti-6.5Al-3.5Mo-1.5Zr-0.3Si Alloy

Laser powder bed fusion(LPBF)is an advanced manufacturing technology;however,inappropriate LPBF process parameters may cause printing defects in materials.In the present work,the LPBF process of Ti-6.5Al-3.5Mo-1.5Zr-0.3Si alloy was investigated by a two-step optimization approach.Subsequently,heat transfer and liquid flow behaviors during LPBF were simulated by a well-tested phenomenological model,and the defect formation mechanisms in the as-fabricated alloy were discussed.The optimized process parameters for LPBF were detected as laser power changed from 195 W to 210 W,with scanning speed of 1250 mm/s.The LPBF process was divided into a laser irradiation stage,a spreading flow stage,and a solidification stage.The morphologies and defects of deposited tracks were affected by liquid flow behavior caused by rapid cooling rates.The findings of this research can provide valuable support for printing defect-free metal components.

基于数据融合的城市轨道交通大客流预警研究

为了降低因节假日或突发时间导致的客流拥堵,促进客运部门快速应对客流波动,提出了一种基于数据融合的城市轨道交通大客流预警方法。通过对客流参数预测信息进行融合,获得符合实际客流瓶颈点的参数信息,以完成轨道交通大客流预警。提出的方法利用动态贝叶斯网络,实现了历史数据与运输服务信息的融合。以城市假期期间五条线路每日客流量为例,通过对比实验验证了提出方法的有效性与准确性。

一种组合电极电容层析成像重建图像融合方法

提出一种基于电容层析成像(ECT)组合电极重建图像及层次聚类的ECT图像融合方法。首先,构建组合8电极及12电极ECT阵列传感器,研究2种组合传感器激励测量方案;其次,采用层次聚类法对两种组合电极ECT重建图像进行聚类,从而消除重建过程中产生的伪影;最后,将聚类后的2种图像再进行小波融合重建。实验结果表明,相较于原始重建图像,所提方法重建图像平均相对误差最小,最大降幅达到23.4%,平均相关系数最大,最大增幅达到10.56%。

基于PCNN图像融合的ECT多相流测量系统设计

针对ECT在多相流测量过程中存在的成像质量不高、精度低、稳定性差等问题,提出并实现了基于PCNN图像融合的ECT多相流测量系统。系统采用模块化设计,主要包括阵列电极传感器模块、电容采集电路模块、上位机成像软件模块。文中设计改进的网络模型,将Tikhonov和Landweber算法的成像结果作为输入进行双通道融合成像,改善了传统算法存在的伪影、模糊的现象。实验结果表明:应用尼龙棒模拟气固两相流对系统进行测试,系统对于微小电容的测量具有良好的稳定性和准确性,通过系统的成像结果可以准确地分辨管道内不同介质的位置,其中对于流型的预测成功率为100%,相含率的测量误差在1%以内。

多尺度融合与动态自适应图的公交客流预测模型

公交客流预测是公共交通规划和管理中的重要问题。虽然时空图卷积在地铁客流预测任务中获得了很好的预测效果,但是面对公交更复杂的线路、大规模的节点数据,现有的基于图卷积的空间建模方法将带来巨大的空间内存消耗。同时,公交客流量短时间范围内更可能受到瞬时交通状况的影响。为了解决这些挑战,提出了一种多尺度融合和动态自适应图的公交客流预测模型(MFDAG)。该模型融合客流、时刻和周信息以增加数据的特征维度,用动态自适应图的方法来学习不同站点之间的关系。进一步提出了一种多尺度融合传播的方法来表示复杂的空间依赖关系,同时设计了一种多尺度卷积传播的方法来学习不同尺度的时间依赖关系。在两个真实的客流数据集上进行了实验,并与其他交通预测方法进行了比较。实验结果表明,所提出的多尺度融合和动态自适应图的公交客流预测方法具有更高的预测准确度。

基于激光雷达探测的飞机尾流融合预测方法

为提高基于激光雷达的飞机尾流探测反演精度,根据扫描出的径向风速数据,建立了基于Kolmogorov结构函数的大气背景湍流耗散率(EDR)估计方法;然后基于两阶段消散预测模型,计入湍流对尾流消散过程的影响,实现基于历史探测数据的尾涡强度环量和涡核运动趋势的预测;通过环境探测数据与预测模型的结合,提高尾涡特性参数的反演精度。研究表明,相较于反演算法,采用本文模型预测的尾涡轨迹在径向距离和角度的精度上分别提高了59.5%、64.8%。

基于幅度和相位融合的微波两相流测量系统设计

针对油井开采过程中需要对各相含量进行精确预测以调整开采策略的实际问题,设计了一种基于微波法的油水两相流检测系统。该系统利用微波在不同介质中透射能力的差异,构建了一套包含微波信号源、功率放大器、功率分配器、检波器和STM32F103ZET6核心板的硬件电路系统,通过编写AD采集和串口通信的软件代码,来接收检波器端幅度和相位数据。在数据处理方面,分别对幅度数据、相位数据和幅度-相位融合数据采用BP神经网络的方法预测含水率。实验表明:在使用融合数据时,预测准确度可以提高至96.33%,取得了较理想的效果。

模块知识融合反应工程大作业设计的思考

化学反应工程是高校化学工程专业的重要核心课之一,致力于研究化学动力学并根据反应特点设计合适的反应器。课程教学团队根据多年教学经验,将课程内容分为理想反应器、非理想流动、非均相催化和工业反应器四个模块,并针对非理想流动与非均相催化模块的知识关联较弱而导致学生无法综合运用所学知识解决复杂反应工程问题的现象,设计了基于模块知识融合的课程大作业。本文阐述了课程大作业对加强模块间知识关联的重要作用,梳理了课程大作业的设计和实施要点,并以催化剂氧空位测量的课程大作业为例,介绍如何通过非理想流动和非均相催化模块知识的融合提升学生解决复杂工程问题的能力,最后总结了课程大作业实施过程中的共性问题并提出了改进建议。

基于VMD-BP-GA模型的脆弱航段船舶短时交通流预测

【目的】针对繁忙航段船舶交通流易受外界环境扰动的难题,提出一种可用于识别船舶交通流脆弱性的预测模型,旨在通过脆弱性辨识,确定最薄弱的航段。【方法】首先采用变分模态分解(VMD)模型将船舶交通流参数序列分解为多个模态分量,然后结合反向传播神经网络(BP)和遗传算法(GA),通过构建约束模型并不断更新各个分量的中心和带宽,实现单个分量的预测,通过应用VMD-BP-GA模型对船舶交通流进行精准预测,并验证其合理性和有效性。【结果】在繁忙航段,本研究提出的VMD-BP-GA模型精准预测船舶交通流脆弱性的方法,相较于传统模型表现出更低的预测误差值,其中在航段流量预测方面,本研究模型的平均绝对误差(MAE)最低达到2.095%,均方根误差(RMSE)最低达到2.610%,平均百分比误差(MAPE)最低达到2.114%;在航段密度预测方面,本研究模型的MAE、RSME、MAPE最低分别为0.129%、0.162%、2.112%;并实现了时空两个维度的船舶交通流预测。【结论】本研究模型成功实现对船舶交通流脆弱性的识别和最薄弱航段的确定,具有高效的预测性能,能够精准并快速地预测船舶交通流,可为船舶通航安全保障提供了理论和实践指导。