基于代理模型的变复杂度方法在板料成形优化中的应用

板料成形优化中存在一些缺陷问题,如一步法与优化算法相结合时精度低、增量法与优化算法直接结合时效率低以及通过传统构造代理模型的方法构造高精度的代理模型时需要大量增量法模拟结果。为了充分发挥一步法和增量法各自的优点,提出采用变复杂度方法通过较少次数的试验样本点数据先在一步法和增量法间建立一个差值补偿响应面模型。通过一步法和差值补偿响应面模型构造新的试验样本点,在新构样本点和原有用增量法计算的样本点数据基础上建立移动最小二乘法代理模型,建立的移动最小二乘法代理模型精度和效率都较高。利用粒子群优化算法对移动最小二乘法代理模型进行优化求解。将该方法应用到了汽车某内板的成形性优化中,优化的结果显著地提高了板料的成形性。算例表明,该方法具有较高的精度和较强的工程实用性。     

计算机辅助测试序列优化

为自动获取平均测试费用最少,且能快速实现系统故障检测与隔离的最优测试序列,提出了一种准多步前向搜索算法。该算法结合了一步前向搜索和多步前向搜索的基本特点,并以信息增益为启发策略,获得了计算精度和复杂度的权衡。基于该算法,开发了一个测试序列优化工具,通过对大量实例的计算,证明该工具可以在实用中取得满意结果,且该工具可以用作自动测试设备和便携式维修辅助软件的核心诊断推理机。     

Study on one-step simulation for the bending process of extruded profiles

Traditional one-step approaches are either based on the membrane element in which bending effects cannot be considered or based on the simplified shell element in which the initial blank is flat and the curvatures are kept unchanged in the one-step iterative computation. They are not suitable for the bending process of extruded profiles. In order to inspect the formability and possible forming defects in profile bending for preliminary designs quickly and to provide an efficient finite element computation for process analysis, a new one-step approach is presented. In this approach, instead of the traditional simplified shell element, the rotation-free basic shell triangular element is introduced to consider bending effects and make the new one-step approach suited to the simulation of profile bending. On the basis of the presented one-step approach, an in-house program named profile bending simulation-one step (PBS-ONESTEP) has been developed to simulate stretch bending of aluminum extrusions and three-point bending of stainless steel extrusions. The algorithm for initial guess solution of extruded profiles is described. Sliding constraint and the penalty method are adopted to treat contacts for the two numerical examples, respectively. The numerical results of PBS-ONESTEP simulation are compared with those of ABAQUS/EXPLICIT incremental analyses and experiments.     

An efficient one-step lookahead A* algorithm for PM-CT scheduling problems

In this paper, we present an improved one-step lookahead A* algorithm for the scheduling problem for parallel-mode cluster tools, where n wafer types are simultaneously processed on a cluster tool with m chambers. Specifically, we suggest a refined cost function, priority-based node selection, N-chamber cycle detection to identify deadlock situations, and a one-step lookahead method to reduce the number of unsafe states generated. Using numerical experiments, we show that the proposed algorithm leads to a significant reduction in the number of explored states and execution time necessary to find an optimal solution.     

一步逆成形法中边界等效阻力的自适应算法

在一步逆成形法中,如果最终构形只包括零件型面,为了得到合理的毛坯形状,需将压边圈的摩擦阻力以及拉深筋的阻力等效到零件边界线的节点上,根据实际经验给出等效阻力的大小。但是等效阻力的大小并不是均匀分布的,如果视为均匀分布,势必造成较大误差,如果分段手动给定,则效率太低,这里提出一种边界等效阻力的自适应算法。算法首先通过线弹性反向变形求得零件在无边界等效阻力作用下的毛坯外轮廓线,假设边界等效阻力的大小与该外轮廓线与零件边界线相应单元边之间的距离成正比,从而实现等效阻力的自适应分布。并以两个典型冲压件为例,通过比较等效阻力自适应算法、等效阻力均布算法、利用商业化软件Fastform分段手动给定阻力的大小和多次试验这四种不同方法得到的零件毛坯外轮廓线,验证了这一算法的有效性和便捷性。     

One-step FEM-based evaluation of weld line movement and development of blank in sheet metal stamping with tailor-welded blanks

Currently, more and more tailor-welded blanks (TWBs) are used in the automobile industry. It is very important to locate the weld line and predict its movement during the forming process. The initial weld line can be predicted by one-step finite-element analysis according to the desired weld line in the final workpiece. Meanwhile, weld line movement during the deformation process can be evaluated in advance. In this paper, the procedures of finite-element analysis of one-step FEM with TWBs are established. The contact between tool and blank and the effect of restraining forces under the blankholder due to drawbead and blankholder pressure are considered as well. Forming limit diagram (FLD) is used to show not only the tendency of reduction in thickness and fracture but also of increase in thickness and wrinkle. Hydraulic controlled pads used to clamp the weld line during the deep drawing process are simplified as static external force to eliminate the movement of the weld line. In order to speed up and ensure the convergence of Newton-Raphson iterations, energy-based 3D mesh mapping algorithm is introduced to obtain the initial solution. The above-mentioned methods have been implemented in the authors’ in-house one-step FEM code InverStamp. A rectangular cup drawing case demonstrates that this approach can be easily implemented to evaluate weld line movement and develop initial blank in sheet metal stamping with tailor-welded blanks.     

基于改进正则法的ECT图像重建算法

电容层析成像图像重建是一个典型的病态问题,其解是不稳定的。为获得良好的重建效果,需要采用既保证解的稳定性且又能提高重建图像质量的算法。本文提出了一种新的图像重建算法。在分析标准Tikhonov正则法的基础上,针对ECT逆问题的病态性进行改进,并推导出两步图像重建算法：第一步利用标准Tikhonov正则法的计算值获得权矩阵的估计;第二步采用本文所推导的改进Tikhonov正则法获得最终的重建图像。数值实验表明,该算法所获得的图像重建质量得到了明显的提高,且该算法无需迭代,保证了算法实时性。     

用有限元法进行复杂结构散热分析方法浅析

随着CAE技术的发展和完善,CAE技术在各行业得到了广泛的运用,但对于复杂的结构件（机箱、机柜和吊舱等）,如果进行详细建模,用一般的硬件条件难以一次性地完成分析计算。文中探讨了一种针对复杂结构进行散热分析的方法,即先简化模型,利用简化模型的计算结果作为局部结构的环境条件,对局部结构进行详细建模,从而得到较为精确的温度分布。并用实例做了对比,得到了较好的效果。     

Estimating the time position of pulses in seismic observation systems based on Markovian filtering

Obtaining the estimates of time position of pulse signals caused by human steps and detected by a seismic observation system is considered. The sequence of pulse appearance moments is described by the Markovian process. The estimates are obtained using nonlinear Markovian filtering. A method for fast calculating a one-step likelihood function without using its approximation is proposed. Statistical modeling results of experimental investigating the synthesized algorithm are presented.     

自适应退火遗传算法的起重机箱形梁优化

为克服标准遗传算法的固有缺陷——停滞和早熟现象,将具有较强局部收索能力的模拟退火算法融入其中,对适应函数进行退火拉伸,对接受算子进行退火处理,同时加入自适应机制来改进标准遗传算法的杂交率和变异率,尤其对变异率的调整,使其既能根据个体适应值的大小进行自适应修正,也能随进化状态的改变而改变,从而增强了算法摆脱局部最优解的能力.以最终形成了自适应退火遗传算法进行起重机主梁优化.经实例验证:与原标准遗传算法相比,在保证收敛结果不变的情况下,收敛速度和全局收敛性都得到了较大提高.     

神经网络在制造单元构建中的研究与应用

分析了ART1神经网络用于制造单元设计的天生缺陷，提出了两条改进途径：以模糊C均值算法对机床一零件矩阵分类问题进行预处理，以提高分类精度；通过修改模式向量的计算方法来克服保存在网络中的模式向量比较稀疏的情况。改进的ART1算法克服了标准ART1算法的不足，成为一种实用有效的制造单元设计方法。设计了新的算法流程并基于相似系数比较尺度在MATLAB软件平台上进行了算法仿真，与前人的研究结果相比，新算法产生了较好的分组效率。     

基于红外光谱技术的沥青质量现场检测

从沥青红外谱图与标准谱图的相似度出发,通过对红外谱图计算PCA得分和峰面积,再通过欧式距离、马氏距离、余弦相似度、皮尔森相似度4个方法,对入库样本与标准谱图进行比对;根据模型结果与分析,选定吸光度PCA得分的余弦相似度和皮尔森相似度两个指标及合格阈值作为基质沥青质量监测的标准,并将算法代码部署在沥青标准光谱分析系统中,实现待测谱图的输入、PCA得分计算、质量是否合格的结果输出。     

Solution of dynamic frictional contact problems using nondifferentiable optimization

This article is devoted to the development, implementation and application of variational inequalities (VI) to treat the general elastodynamic contact problem. Unlike existing VI algorithms, which rely on the decomposition of the physical problem into two sub-problems, the current study treats the general VI expressions using one-step approach to identify the candidate contact surface and contact stresses. This is accomplished using nondifferentiable optimization algorithm, through a sequence of mathematical programming problems. The dynamic VI formulations are solved using generalized-α time integration scheme. The selected time integration parameters reduce significantly the spurious high-frequency modes, which persist in the traditional Newmark method. In order to demonstrate the versatility and accuracy of the proposed VI and nondifferentiable optimization algorithm, a number of numerical examples are examined. The results show a significant improvement compared with the existing solution techniques.     

改进Elman神经网络在短期热负荷预测中的应用

热电厂的短期热负荷预测在城市集中供暖中起着至关重要的作用,直接影响热电厂的经济效益和热能利用率。电厂的短期热负荷一般采用神经网络预测模型进行预测,而BP神经网络应用最为广泛。Elman神经网络算法在BP神经网络基础上加入了承接层,作为一步延时算子,实现记忆能力,使系统具备适应时变能力,增强系统全局稳定性。但Elman神经网络算法模型的构造依然需要大量样本的支撑,而且输入层的变量多,导致预测时间依然很长,收敛速度慢。该文在Elman神经网络预测前,进行了相关系数预处理和对样本中异常值的平均化预处理,通过数据归一化运算,使Elman神经网络输入层变量大幅减少。仿真实验表明,改进的Elman神经网络算法使预测模型快速寻优,减少预测时间的同时明显提高预测精度。     

Conversion of wireframe to ACIS solid models for 2 1/2-D engineering components

Wireframe models need to be converted to solid models as the latter tend to become more widespread, consistent and affordable. A conversion algorithm for manifold objects was defined to obtain ACIS, thede facto standard B-rep model from a neutral IGES wireframe model and was implemented in C++ in successive stages as follows. All potential surfaces are found first. The entities lying on the same surface are examined in order to form closed profile-loops. Loop nesting is examined according to a ray casting algorithm. Adjacency and nesting of loops is used in order to assign traversal directions to all edges according to the Möbius rule. Invalid profiles (those containing edges belonging to more than two loops) are searched for and successively removed. The final test is that of edge type. If the majority of edges turn out to be concave, the traversal sign of all edges is reversed, so that the body is solid in empty space and not the other way around. For engineering components the algorithm adequately resolves all singular cases.     

图像端到端感知质量的快速估测方法

本文介绍了一种新型的利用图像活动性(IAM)快速估计水下图像端到端体验质量的方法。首先引入结构相似度(SSIM)作为图像感知服务质量(PQo S)参数评价图像质量,将图像初始活动性(IAM)作为区分图像内容的本征参数;随后基于质量向量(QV)的概念,分别分析了非压缩图像的结构相似度,图像初实活动性与无条件丢失概率(SSIM-IAM.-ulp)之间的联合特性,以及压缩图像的结构相似性,压缩率与无条件丢失概率(SSIM-IAM.-ulp)之间的联合特性。最后,在上述联合特性的基础上提出了劣化图像SSIM的预测算法。测试实验证明,该类算法有较高的预测准确率,预测误差最低可达0.8%。     

基于广义预测控制算法的水槽液位控制系统

为提高液位系统控制的可靠性和安全性,引入了广义预测控制算法。首先介绍了广义预测控制的基本算法,包括其基本理论、算法的优点及一些重要参数的选取标准。然后以水槽液位控制装置为控制对象建立了一个控制系统的数学模型,并对应用广义预测控制算法的该模型进行了仿真验证。仿真结果表明,所提出的算法具有良好的动态响应性能和快速跟踪设定值,并具有较好的控制效果。无论系统是否存在模型失配,只要控制系统是稳定的,就不会有静差。     

用改进的CT法估算离散点云的法矢和曲率

逆向工程中一个基本而关键的步骤是基于离散数据点估算法矢和曲率.针对目前常用估算方法CT法的缺点进行改进,提出以传统CT法估算的法矢为初值,通过迭代的过程提高计算精度.实验结果表明,所提出的改进算法不但简单可行,而且能够大大提高计算精度.     

Third‐harmonic generation imaging of breast tissue biopsies

We demonstrate for the first time the imaging of unstained breast tissue biopsies using third‐harmonic generation (THG) microscopy. As a label‐free imaging technique, THG microscopy is compared to phase contrast and polarized light microscopy which are standard imaging methods for breast tissues. A simple feature detection algorithm is applied to detect tumour‐associated lymphocyte rich regions in unstained breast biopsy tissue and compared with corresponding regions identified by a pathologist from bright‐field images of hematoxylin and eosin stained breast tissue. Our results suggest that THG imaging holds potential as a complementary technique for analysing breast tissue biopsies.     

一种改进的微粒群算法--多步长微粒群算法

首先分析了标准微粒群算法及影响全局和局部搜索能力的主要因素。在此基础上，构造了一种改进的微粒群算法——多步长微粒群算法。实例计算表明，该算法比标准微粒群算法具有更好的性能。