基于Gurtin变分原理的一种逐步积分法

基于Gurtin变分原理,利用经过空间离散后的只含单重卷积形式的泛函,在局部时间域上采用初位移、初速度和末位移、末速度并加入一种非时间步参数的插值函数形式对时间域进行离散,给出了一种逐步积分法。文中通过对动力计算方法的稳定性研究,选取了合适的非时间步参数。精度分析和数值算例表明本文方法是一种可以应用于结构动力响应分析并具有较高精度的方法。     

Rapid and Accurate Contact Determination between Spline Models using ShellTrees

In this paper, we present an efficient algorithm for contact determination between spline models. We make use of a new hierarchy, called ShellTree , that comprises of spherical shells and oriented bounding boxes. Each spherical shell corresponds to a portion of the volume between two concentric spheres. Given large spline models, our algorithm decomposes each surface into Bézier patches as part of pre-processing. At runtime it dynamically computes a tight fitting axis-aligned bounding box across each Bézier patch and efficiently checks all such boxes for overlap. Using off-line and on-line techniques for tree construction, our algorithm computes ShellTrees for Bézier patches and performs fast overlap tests between them to detect collisions. The overall approach can trade off runtime performance for reduced memory requirements. We have implemented the algorithm and tested it on large models, each composed of hundred of patches. Its performance varies with the configurations of the objects. For many complex models composed of hundreds of patches, it can accurately compute the contacts in a few milliseconds.     

Optimization of stress modes by energy compatibility for 4‐node hybrid quadrilaterals

Optimal hybrid stress quadrilaterals can be obtained by adopting appropriate stresses and displacements, and satisfying the energy compatibility condition is shown to be an ultimate key to obtaining optimal stress modes. By using compatible isoparametric bilinear (Q₄) displacements and 5‐parameter energy compatible stresses of the combined hybrid finite element CH(0‐1), a robust 4‐node plane stress element ECQ₄ is derived. Equivalence to another hybrid stress element LQ₆ with 9‐parameter complete linear stresses based on a modified Hellinger–Reissner principle is established. A convergence analysis is given and numerical experiments show that elements ECQ₄/LQ₆ have high performance, i.e. are accurate at coarse meshes, insensitive to mesh distortions and free from locking. Copyright © 2003 John Wiley & Sons, Ltd.     

Three‐dimensional finite element computations for frictional contact problems with non‐associated sliding rule

This paper presents an algorithm for solving anisotropic frictional contact problems where the sliding rule is non‐associated.The algorithm is based on a variational formulation of the complex interface model that combine the classical unilateral contact law and an anisotropic friction model with a non‐associated slip rule. Both the friction condition and the sliding potential are elliptical and have the same principal axes but with different semi‐axes ratio. The frictional contact law and its inverse are derived from a single non‐differentiable scalar‐valued function, called a bi‐potential. The convexity properties of the bi‐potential permit to associate stationary principles with initial/boundary value problems. With the present formulation, the time‐integration of the frictional contact law takes the form of a projection onto a convex set and only one predictor–corrector step addresses all cases (sticking, sliding, no‐contact). A solution algorithm is presented and tested on a simple example that shows the strong influence of the slip rule on the frictional behaviour. Copyright 2004 John Wiley & Sons, Ltd.     

基于深度学习直流闭锁判断的高风险连锁故障快速搜索

大容量直流输电系统的快速发展及多馈入直流系统的出现,增加了系统发生连锁故障的风险。建立了基于深度学习的直流闭锁快速判断模型,提出一种高风险连锁故障快速搜索方法。直流闭锁快速判断模型以网络结构和故障位置相关的稳态特征量作为输入,利用堆叠降噪自动编码器(SDAE)提取输入的高阶特征,可以快速判断交流故障是否导致直流闭锁。根据交流故障后正常线路的负载率情况衡量该故障的影响程度,进而得到线路停运风险;搜索过程中采用深度优先策略,以高停运风险为搜索方向,优先获取故障概率较高的交直流连锁故障。实际电网仿真结果表明,所提方法能够快速给出高风险交直流连锁故障的传播路径和故障概率,可用于在线安全预警和防控决策。     

基于CVMD和DBO-SVM的光纤周界安防信号识别方法

为降低光纤周界安防信号中噪声对分类结果的影响,提升信号分类的准确率和运行效率,提出一种融合了相关变分模态分解(Correlation Variational Mode Decomposition, CVMD)、蜣螂算法(Dung Beetle Optimizer, DBO)和支持向量机(Support Vector Machine, SVM)的分类方法。利用CVMD去除原始信号中的噪声分量,并提取去噪后信号的能量、能量熵和峭度作为特征向量。采用DBO算法优化SVM,得到最佳惩罚因子和核函数参数,并构建DBO-SVM分类模型。搭建了基于相位敏感光时域反射(Φ-OTDR)技术的周界安防系统,采集了攀爬、敲击、踩踏和无入侵四类信号。实验结果表明,CVMD-DBO-SVM的分类准确率相比CVMD-PSO-SVM和CVMD-GA-SVM更高,达到了98.75%,同时运行时间更短,综合性能最优。     

双向时域特征流盲去运动模糊方法

便携式成像设备在日常的生活中已经无处不在,但是因为拍摄时的抖动或者场景中的快速运动物体,所拍摄的图像或视频经常存在模糊现象,造成图像重要细节丢失。为了将模糊的视频图像恢复到清晰的状态,结合近年来的研究热点——生成对抗网络,本文提出了一种新颖的端到端的双向时域特征流盲去运动模糊方法。该方法充分利用时空连续特征信息,在三连帧图像间建立双向的时域特征传输通道。多阶段自编码去模糊网络结构和并行编码-混合解码融合方案能够融合三连帧图像多通道内容信息,并恢复出更加清晰的视频图像。实验结果表明,在不牺牲较大时间代价前提下,本文提出的方法在传统的质量评价指标（峰值信噪比和结构相似性）和视觉质量上均优于现有的去模糊算法。     

基于EMD距离的稀疏自编码器

KL散度在机器学习领域被广泛地用于模型损失函数之中来度量分布的距离。在稀疏自编码器中,KL散度被用作损失函数的惩罚项来度量神经元输出与稀疏参数的距离,使得神经元输出趋近稀疏参数,从而抑制神经元的激活以得到稀疏编码。在WGAN中,Wasserstein距离被用于解决GAN的梯度消失和模式塌陷问题,使得GAN的训练更加稳定。得益于Wasserstein距离在GAN中的成功应用,提出了基于EMD距离的稀疏自编码器SAE-EMD。实验结果表明,相比于使用KL散度与JS散度,使用EMD距离作为惩罚项的稀疏自编码器可以使得真实样本与重构样本之间的重构误差减小,并且随着惩罚参数的增大,编码更加稀疏。     

基于卷积自编码器的座椅电机故障诊断

故障数据的缺失一直是制约设备故障诊断发展的重要因素,现有研究通过刻意损坏设备的方法来采集故障数据。为实现座椅电机的无损故障诊断,文章对座椅电机的故障机理进行分析,确定可能发生的故障类型,通过在座椅电机表面粘贴微型喇叭并播放故障声音,来模拟故障的发生。在自编码器系统的基础上,引入卷积操作,使用卷积层代替全连接层,通过输入数据维度、卷积核的尺寸和数量以及池化、正则化等操作对模型结构进行调整。采用IDMT Isa Electric Engine数据集作为源域数据,对模型进行预训练。使用迁移学习方法将源域中已经学习到的数据分布迁移到座椅电机故障诊断任务中,并与各类模型检测结果进行对比。结果显示,文中方法在召回率保持1.00的情况下,曲线下面积达到0.86,检测结果可靠,具有实际应用价值。     

测量噪声方差未知的多传感器组合导航集中融合算法

目前,多传感器组合导航系统的信息融合方法是建立在测量噪声方差已知的基础上,然而测量噪声方差会随着内部及外部的干扰而发生变化。为此,本文首先将基于变分贝叶斯逼近的自适应卡尔曼滤波(variational Bayesian approximation based adaptive Kalman filter, VB-AKF)从单一组合导航系统扩展到多传感器组合导航系统;然后,提出了多传感器组合导航系统的两种集中融合算法,即基于VB-AKF的增广式集中融合算法及基于VB-AKF的序贯式集中融合算法,以解决测量噪声方差未知情况下的多传感器组合导航的信息融合问题;最后,通过SINS/GNSS/CNS/ADS多传感器组合导航系统对上述算法进行了仿真验证。实验结果表明,本文所提两种算法滤波精度相同、且接近于测量噪声方差已知情况下的理想集中融合算法(ICKF)。在整个仿真时段内,相对于传统集中式卡尔曼滤波器(TCKF)及具有容错功能的联邦卡尔曼滤波算法(FT-FKF),本文算法可提高位置精度分别为32%和90%、提高速度精度分别为38%和71%。