随机结构数值模拟分析的神经网络法

在随机结构分析中,蒙特卡洛方法作为随机数值模拟方法,为问题提供了最为直观和精确的解答,但计算量大、效率低下的缺点大大降低了方法的实用性.研究在蒙特卡洛方法中引入人工神经网络,仅进行少量确定性分析,训练后即可模拟确定性有限元求解器,用神经网络的快速泛化映射取代蒙特卡洛法中的大量确定性有限元分析.算例结果显示,提出的蒙特卡洛-神经网络法可将蒙特卡洛法的计算效率提高几十至一百倍,计算精度令人满意,是一种有潜力的随机结构实用分析方法.     

基于多分支BP神经网络的结构系统辨识

结构动力学系统是一个多输入多输出系统,建立能精确体现结构动力特性的辨识模型是实现结构高品质振动控制和关系到控制是否有效的关键。提出一种多分支BP神经网络辨识模型,将影响结构动力反应的结构状态变量和地震动输入分别作为模型的分支输入来进行辨识,提高学习效率及预测精度,并利用训练好的模型预测结构在不同地震波输入下的动力反应,验证模型的泛化能力。数值分析结果表明,用所提出的多分支BP网络模型对结构动力学系统进行动力特性辨识时能达到较高的精度,而且预测精度也很高。     

H∞-学习神经网络的非线性系统识别

基于H∞-滤波算法提出了一种有效的回归神经网络学习方法。该方法能有效解决系统模型和噪声统计特性存在不确定的问题。在利用神经网络进行非线性系统识别中，该方法对任意的网络结构能进行整体优化训练，具有良好的泛化能力。在非线性结构系统识别中验证了该方法的可行性与有效性。     

NEW REGULARIZATION BY TRANSFORMATION FOR NEURAL NETWORK BASED INVERSE ANALYSES AND ITS APPLICATION TO STRUCTURE IDENTIFICATION

The present authors have been developing an inverse analysis approach using the multilayer neural network and the computational mechanics. This approach basically consists of the following three subprocesses. First, parametrically varying model parameters of a system, their corresponding responses of the system are calculated through computational mechanics simulations such as the finite element analyses, each of which is an ordinary direct analysis. Each data pair of model parameters vs. system responses is called training pattern. Second, a neural network is iteratively trained using a number of training patterns. Here the system responses are given to the input units of the network, while the model parameters to be identified are shown to the network as teacher data. Finally, some system responses measured are given to the well-trained network, which immediately outputs appropriate model parameters even for untrained patterns. This is an inverse analysis. This paper proposes a new regularization method suitable for the inverse analysis approach mentioned above. This method named the Generalized-Space-Lattice (GSL) transformation transforms original input and/or output data points of all training patterns onto uniformly spaced lattice points over a multi-dimensional space. The topological relationships among all the data points are maintained through this transformation. The neural network is then trained using the GSL-transformed training patterns. Since this method significantly remedies localization of training patterns caused due to strong nonlinearity of problem, the neural network can learn the training patterns efficiently as well as accurately. Fundamental performances of the present inverse analysis approach combined with the GSL transformation are examined in detail through the identification of a vibrating non-uniform beam in Young's modulus based on the observation of its multiple eigenfrequencies and eigenmodes.     

基于神经网络的桥梁损伤位置识别

损伤位置识别是对大型桥梁结构进行损伤检测的重要一步.以汲水门斜拉桥为背景,对应用神经网络的模式分类技术识别桥面结构损伤位置的方法进行了研究.使用了两种类型的网络,简称动态网络和GA网络,探讨了这一方法的可行性.动态网络的网络结构是根据训练进程而动态地确定的.GA网络是在训练中引进了遗传算法.比较了两种网络对损伤位置的识别效果.结果表明,应用神经网络的模式分类技术对桥梁桥面结构损伤位置识别的方法是可行的.只需要较少的输入数据,两种网络均可产生较好的识别结果.     

用神经网络求解物资运输问题

利用神经网络算法实现了对物资调拨运输问题的优化计算,证明该方法是切实可行的,计算机的模拟结果令人满意。同时也给类似的问题提供了借鉴。     

工程结构损伤的集成神经网络识别研究

从结构损伤识别的实际出发，提出采用基于信息融合理论的集成神经网络技术对结构损伤状况进行识别，即通过结构损伤特征信息的有效组合，用各种子神经网络从不同侧面对结构损伤进行初步识别诊断，然后对识别结果进行决策融合。给出了系统的实现策略和子网络的组建原则。从识别实例中可以看出，此识别方法充分利用了各种特征信息，可以有效地提高识别率。     

用神经网络对内燃机气门间隙的识别

由于内燃机的振动信号较复杂，用一般分类器来识别它的故障，不容易得到满意的效果。本文提出利用神经网络构成的分类器，对内燃机气门间隙进行识别，实验与计算机模拟，得到较满意的结果。     

神经网络在轨道动力学中的应用

本文中研究了神经网络在列车轨道动力学中的应用。列车轨道系统的各种参数众多,关系复杂,用模型分析出入较大。为了利用试验数据,分析列车轨道动力特性,本文提供了一种应用神经网络的分析方法。结果表明,用神经网络整理分析试验数据是有效的     

基于结构响应统计特征的神经网络损伤识别方法

提出一种采用结构动态响应的统计特征作为损伤指标的神经网络损伤识别方法,并对其进行了数值模拟和实验验证。首先,通过敏感性分析,分析了采用结构动力响应的统计特征作为损伤指标的可行性;然后数值模拟了一三跨连续梁采用结构位移方差作为损伤指标的神经网络损伤识别过程,其结果表明,经过训练的神经网络可以准确的识别出单损伤和多损伤工况中的损伤位置和损伤程度;最后进行一组两端固定的简支梁模型实验来验证所提出损伤识别方法的有效性。实验结果表明,对于单损伤工况,神经网络可以准确地识别出结构中损伤位置和损伤程度,对于双损伤工况,神经网络可以准确地识别出损伤位置,而损伤程度识别略有偏差。最后得出结论,采用结构动力响应的统计特征作为损伤指标的神经网络损伤识别方法是可靠有效的。     

基于模糊神经网络的数据融合结构损伤识别方法

为了有效利用结构健康监测系统中的多源传感器数据信息,提高损伤检测与评估的识别正确率,该文通过构造模糊神经网络分类器,提出了一种基于模糊神经网络的数据融合损伤识别方法并将之应用于结构健康诊断中。它先通过数据预处理,提取原始响应信号中的特征参数,接着将之作为模糊神经网络的输入,构造模糊神经网络模型进行识别决策,最后运用数据融合算法,计算出数据融合后的决策结果。为了验证所提方法的有效性,通过一个7自由度的建筑模型,分别用单一模糊神经网络决策器和数据融合损伤识别方法进行了损伤识别和比较。研究结果表明:该文所提方法比单一决策结果更准确、可靠。     

人工神经网络-研究制冷系统的新方法

本文介绍了应用于工程研究中的MLP、RBF神经网络的结构及学习训练算法，结合实例详述了将人工神经网络技术应用于制冷系统研究的各种途径，分析了神经网络应用的难点及解决办法，并指出人工神经网络技术在制冷系统研究中的应用方向。     

基于神经网络的损伤构件及损伤程度识别

确定损伤构件及其损伤程度是分阶段损伤识别的最后一步,同时又是进一步制定结构安全运行决策的前提和基础。研究了在确定了结构损伤区域的条件下,应用反向传播(BP)神经网络同时实现对具体损伤构件及其损伤程度识别的方法。探讨了针对上述神经网络训练数据的构造和训练策略。应用提出的方法对汲水门斜拉桥桥面结构进行了损伤识别仿真模拟。基于模态参数对损伤的灵敏度分析,选取了12个自振频率和损伤区域附近的6个振型分量作为构造网络输入的基本数据。网络的输出向量同时指示了损伤构件及其损伤程度。就模拟的10种损伤情况,当损伤程度达到60%以上时,有9种实现了正确的构件识别,半数以上给出了可以接受的损伤程度描述。     

基于RBF神经网络的复合材料层合壳荷载识别

针对有限元逆分析方法进行荷载识别的大计算量的缺陷,以及鉴于传统的BP网络的速度慢和局部极小值问题,该文提出了将有限元方法与径向基函数(Radial Base Function,简记为RBF)神经网络结合对受集中载荷作用的壳体结构进行荷载识别。通过有限元方法计算出压电元件的集聚电荷,以该电荷来构建训练样本对网络进行训练,再将没有进行训练的电荷数据送入到训练好的RBF神经网络进行预测,实现对壳体结构荷载的作用位置和大小的评估。最后给出了对壳体结构荷载识别的算例,结果表明该方法计算速度快、精度高、具有较好的应用前景。     

应用于磨床砂轮磨钝及故障检测的神经网络研究

磨床在磨削过程中砂轮会逐渐变钝。当它变钝后,需要进行修快。针对砂轮变钝时间的不定性,利用神经网络进行砂轮监控,以便于充分把握修快时机,达到及时、经济的修快。同时对砂轮意外碰撞、破碎进行监控。     

TOOL WEAR MONITORING UNDER DRY AND WET MACHINING

External turning operations for steel 37 (AISI 1025) tubes were performed on a center lathe using high-speed steel tools. The workpiece was machined orthogonally using a wide range of cutting speeds for both dry and wet cutting conditions. The tool wear and the constants of the tool life equation (C, n) were studied by changing the cutting speed and the wear criterion. Comparisons between dry and wet cutting for different cutting speeds were performed, and the constants of the tool life equation were studied under different wear criterion values. The economy of dry and wet cutting was also studied under different wear criterion values. As a result of using cutting fluids, percentage increase in the constant C reached 18%, whereas percentage decrease in the constant n reached 7% at 0.9 wear criterion, which leads to a percentage increase in the tool life of about 214%.     

一种弧长参数曲线及其计算机实现

本文提出了一种弧长参数曲线的数值算法，这种算法具有收敛迅速，迭代过程自动确定等优点。     

GENETIC ALGORITHMS AS AN APPROACH TO OPTIMIZE REAL-WORLD TRUSSES

Genetic algorithms, a search technique which combines Darwinian ‘survival-of-the-fittest’ with randomized well structured information, is applied to the problems of real-world truss optimization. In this work a population of binary strings or ‘chromosomes’, which represent the coded truss design variables, a ‘fitness’ as a ranking measure of the adaptability to the environment, selection criteria and mechanical natural operators such as crossover and mutation are used to improve the population, so that over the generations the genetic algorithm gets better and better and at the end of the convergence, a ‘rebirth’ of the population is used to improve the usual process. An overview of the genetic algorithm will be described, continuing the rebirth effect; then, the chromosome representation of trusses is exposed. Afterwards, the objective scalar function is defined taking into account that it seems reasonable in real world to optimize trusses in minimum weight trying, at the same time, to use the minimum number of cross-section types obtained from the market. It also seems reasonable to have the possibility to change the shape of the conceptual design, moving some joints. To simulate nearly real conditions, several load cases, constraints in the elastic joint displacements, ultimate tensile and elastic and plastic buckling in the bars have been taken into account. A hyperstatic 10 bars truss is subjected to a deep analysis in different situations in order to evaluate with other authors when possible as truss optimization with two criteria and buckling effect has not been found in specialized literature. A 160-bar transmission tower is also optimized.     

EFFECT OF TOOL WEAR AND TOOL SETTING ON PROFILE ACCURACY OF DIAMOND-TURNED NONFERROUS COMPONENTS

Diamond turning technology has gained great importance in high-precision optical component fabrication. The quality of machined optical surfaces is mainly affected by the machine tool's accuracy, cutting tool's quality, and dynamic machining effects. This study investigated the effects of cutting tool conditions and tool set-up error on the surface distortion. Controlled cutting tests were performed on a two-axis diamond turning machine. Spherical mirrors with preset tool offset values and tool height values were turned. The relationship among machined form accuracy, tool offset, and tool height was investigated based on experimental and analytical results. The influence of tool wear on machined surface quality was studied. Factors governing uneven wear along the cutting edge in contour machining were analyzed. A spherical surface with a form accuracy better than λ/10 was produced. Methods for minimizing the effect of tool wear are also discussed.     

AN APPROACH TO THE SYSTEMATIC ARRANGEMENT OF WORK MEASUREMENT DATA

A model is.proposed for the systematic arrangement of work element times. This model consists of a grouping of all work element times along the time co-ordinate either in groups of equal size or with the time span of each group increasing with the mean time of each group. The statistical properties in analysing work by means of data thus arranged are investigated. A comparison is madewith the sampling error.of.such,data, the observational error, when the error consists of referring the work elements to the wrong group, and the natural variation of the performance time.