大跨度斜拉桥可靠度评估方法研究

针对大型复杂桥梁结构极限状态方程一般难以显式表达的特点,提出了基于神经网络的大跨度斜拉桥可靠度评估方法．通过Latinhypercube抽样技术对随机参数进行抽样,应用大跨度斜拉桥非线性有限元进行分析．通过对随机抽样的样本数据进行训练,应用神经网络的非线性映射和泛化技术,对大跨度斜拉桥的极限状态方程进行数值模拟．通过极限状态方程对随机变量的偏导数,求解结构可靠指标的优化问题,计算大跨度斜拉桥的可靠指标．结果表明：对于隐式极限状态方程的大跨度斜拉桥可靠度评估问题,本文方法具有较高的计算精度和较好的计算效率;荷载布置方式、作用位置等对斜拉桥可靠指标有很大影响;计入3种几何非线性效应后斜拉桥偏于不安全,其中斜拉索垂度非线性效应的影响最为显著．     

预应力混凝土连续箱梁有限元损伤分析(英文)

为了确定预应力混凝土箱粱的极限承载能力,进行了预应力混凝土连续箱梁的破坏性试验,提出一种有限元损伤分析方法模拟箱梁结构行为.基于实体退化壳单元对箱梁进行模拟,并考虑了结构材料的非线性效应.由于预应力混凝土箱梁中通常存在大量曲线预应力钢筋,提出一种组合单元,并用该单元模拟箱梁中的预应力钢筋,该方法可使结构分析时单元数量大大降低.试验结果和理论分析对比表明:提出的方法能够有效地用于预应力混凝土连续箱梁的破坏性试验计算分析中;研究的旧桥依然存在较大的极限承载能力.     

5ayes统计中两种抽样法的比较分析

拒绝性抽样法和重要性抽样法是贝叶斯统计中两种重要的产生模拟样本的方法．两种方法之间存在一些关联,拒绝性抽样法在做一些形式上的转变后可以看作一种特殊的重要性抽样法．此外,如果将拒绝性抽样法中的包络函数作为重要性抽样法中的提议密度函数,这样新形成的抽样法会比原本的拒绝性抽样法更为有效．     

5ayes统计中两种抽样法的比较分析

拒绝性抽样法和重要性抽样法是贝叶斯统计中两种重要的产生模拟样本的方法．两种方法之间存在一些关联,拒绝性抽样法在做一些形式上的转变后可以看作一种特殊的重要性抽样法．此外,如果将拒绝性抽样法中的包络函数作为重要性抽样法中的提议密度函数,这样新形成的抽样法会比原本的拒绝性抽样法更为有效．     

手写体数字字符识别算法仿真比较研究

采用BP神经网络、原始极限学习机、正则极限学习机和傅里叶变换优化极限学习机算法分别进行手写体数字字符识别仿真实验,通过MINIST数据库中的10 000个手写体数字样本训练神经网络数据传输过程中的参数,用训练所得神经网络参数进行手写体数字识别仿真测试。比较4种算法的仿真效果,BP网络识别效果最好、训练速度最慢,原始极限学习机训练速度最快,正则极限学习机和傅里叶变换极限学习机优化算法识别正确率高于原始极限学习机,但增加了算法的时间复杂度。将该手写体数字字符识别仿真实验用于学生实验教学,可加强学生对神经网络相关知识的学习和理解,提高学生编程和工程应用能力。     

ANSYS软件在砌体结构非线性分析中的应用

利用ANSYS软件建立空间模型,对轴心受压时设置构造柱的砖砌体结构进行了模拟．得到的结果与试验结果较吻合,并优于以往平面有限元的计算结果。说明可以利用ANSYS软件强大的模拟功能对砌体结构进行有限元分析．为目前日益增多的各类砌体结构提供非线性分析依据。     

云计算环境下并行进化神经网络的设计研究

针对提高进化神经网络进化时效性,充分利用神经网络的训练数据,提出一种在云计算Hadoop平台环境下,使用进化算法对BP神经网络的权值和网络结构进行优化,通过分布并行计算,提高进化速度和效率.理论分析和实验结果表明,在数据量较大时,该方法能有效地提高神经网络计算精度.     

ANSYS软件在砌体结构非线性分析中的应用

利用ANSYS软件建立空间模型,对轴心受压时设置构造柱的砖砌体结构进行了模拟．得到的结果与试验结果较吻合,并优于以往平面有限元的计算结果。说明可以利用ANSYS软件强大的模拟功能对砌体结构进行有限元分析．为目前日益增多的各类砌体结构提供非线性分析依据。     

粗糙集和神经网络在文本分类中的应用研究

把粗糙集与神经网络结合,应用于文本分类,可以充分发挥两种方法的优势,取长补短,粗糙集理论可以有效地对样本集进行约简,从而简化了神经网络的结构,减少了网络的训练次数,学习速度和分类精度明显提高,并用仿真实验验证了此方法的有效性.     

“1^∞”型重要极限的应用

重要极限：limx→∞（1＋1／n）^n=e的研究对极限的计算与教学至关重要．此类极限可归结为“1^∞”型不定式极限．为此将利用等价无穷小替换的方法对这一类型极限的计算进行详细讨论，得到计算这一类型极限简便快捷的方法．     

Finite element reliability analysis of slope stability

The method of nonlinear finite element reliability analysis （FERA） of slope stability using the technique of slip surface stress analysis （SSA） is studied. The limit state function that can consider the direction of slip surface is given, and the formulations of FERA based on incremental tangent stiffness method and modified Aitken accelerating algorithm are developed. The limited step length iteration method （LSLIM） is adopted to calculate the reliability index. The nonlinear FERA code using the SSA technique is developed and the main flow chart is illustrated. Numerical examples are used to demonstrate the efficiency and robustness of this method. It is found that the accelerating convergence algorithm proposed in this study proves to be very efficient for it can reduce the iteration number greatly, and LSLIM is also efficient for it can assure the convergence of the iteration of the reliability index.     

Calculation method of ship collision force on bridge using artificial neural network

Ship collision on bridge is a dynamic process featured by high nonlinearity and instantaneity. Calculating ship-bridge collision force typically involves either the use of design-specification-stipulated equivalent static load, or the use of finite element method （FEM） which is more time-consuming and requires supercomputing resources. In this paper, we proposed an alternative approach that combines FEM with artificial neural network （ANN）. The radial basis function neural network （RBFNN） employed for calculating the impact force in consideration of ship-bridge collision mechanics. With ship velocity and mass as the input vectors and ship collision force as the output vector, the neural networks for different network parameters are trained by the learning samples obtained from finite element simulation results. The error analyses of the learning and testing samples show that the proposed RBFNN is accurate enough to calculate ship-bridge collision force. The input-output relationship obtained by the RBFNN is essentially consistent with the typical empirical formulae. Finally, a special toolbox is developed for calculation effi- ciency in application using MATLAB software.     

Evaluation of accelerated deterioration in NAPTF flexible test pavements

Previous research studies have successfully demonstrated the use of artificial neural network （ANN） models for predicting critical structural responses and layer moduli of highway flexible pavements. The primary objective of this study was to develop an ANN-based approach for backcalculation of pavement moduli based on heavy weight deflectometer （HWD） test data, especially in the analysis of airport flexible pavements subjected to new generation aircraft （NGA）. Two medium-strength subgrade flexible test sections, at the National Airport Pavement Test Facility （NAPTF）, were modeled using a finite element （FE） based pavement analysis program, which can consider the non-linear stress-dependent behavior of pavement geomaterials. A multi-layer, feed-forward network which uses an error-backpropagation algorithm was trained to approximate the HWD back- calculation function using the FE program generated synthetic database. At the NAPTF, test sections were subjected to Boeing 777 （B777） trafficking on one lane and Boeing 747 （B747） trafficking on the other lane using a test machine. To monitor the effect of traffic and climatic variations on pavement structural responses, HWD tests were conducted on the trafficked lanes and on the untrafficked centerline of test sections as trafficking progressed. The trained ANN models were successfully applied on the actual HWD test data acquired at the NAPTF to predict the asphalt concrete moduli and non-linear subgrade moduli of the medium-strength subgrade flexible test sections.     

应用人工神经网络（ANN）预测结构的地震响应

提出了基于人工神经网络（ＡｒｔｉｆｉｃｉａｌＮｅｕｒａｌＮｅｔｗｏｒｋｓ）对动力结构进行系统辨识的方法，即应用人工神经网络预测结构地震响应．采用ＢＰ算法的前馈网络（简称ＢＰ网络）对剪切模型结构进行系统辨识．首先用实际地震波及相应的模拟地震响应训练本文提出的ＢＰ网络，然后用“已学会”的ＢＰ网络预测其它地震波激励下的结构地震响应．还讨论了网络拓扑结构、输入单元数等对网络学习和预测的影响．通过本文可以发现，合适的人工神经网络结构能准确地辨识结构动力特性和预测结构动力响应     

Reliability analysis of diesel engine crankshaft based on 2D stress strength interference model

INTRODUCTION For critical structures and applications, requiring a given reliability, it is necessary to account for un-certainties and variability in material properties, loads and geometric tolerances. Using various reliability analysis methods, we can determine the reliability of existing structures or systems; or design new structures or systems with certain desired reliability. Manufacturing processes may involve many factors that have inherent uncertainties or random variations in th…     

随机有限元的直接法

提出了一种不修改通用有限元程序建立的随机有限元方法，即计算机随机产生多组零件尺寸，用通用有限元软件对每组数据进行计算，求出零件危险截面应力等。对应力数据进行检验，若应力服从正态分布、指教分布、威布尔分布、对教正态分布中任一种，强度可靠度可通过公式或数值积分得到。若应力分布不确定，则可靠度用Monte Carlo法求出。变形的可靠度用同样方法解决。用该方法编写的计算机程序，使形状复杂零件的可靠度计算成为可能。     

基于人工神经网络的研究生课程评价模型

在研究生网上课程评价系统收集的大量数据基础上,构建了一种基于人工神经网络的多指标课程评价模型,并将网上调查的结果以连接权的方式赋予该评价模型进行训练。通过实际评价数据的验证,该模型能够准确地按照实际评价的过程进行工作。     

基于响应面法的钢拱桥可靠度分析

大型复杂桥梁结构的极限状态函数都是隐式的非线性函数,采用传统的可靠性分析方法可能无法实现,但利用响应面有限元法可以很好地解决这类问题。利用响应面法对一下承式钢拱桥拱肋、主梁在正常使用极限状态下的轴向应力、竖向位移的可靠度进行了分析,并根据拟合出的各极限状态函数显式表达式得出了相应的可靠度指标,说明响应面有限元法是计算下承式钢拱桥构件可靠度的有效方法。     

Damage identification in beam-type structures using pseudo strain energy density and grey relation coefficient

Based on pseudo strain energy density （PSED） and grey relation coefficient （GRC）, an index is proposed to locate the damage of beam-type structures in time-domain. The genetic algorithm （GA） is utilized to identify the structural damage severity of confirmed damaged locations. Furthermore, a systematic damage identification program based on GA is developed on MATLAB platform. ANSYS is employed to conduct the finite element analysis of com- plicated civil engineering structures, which is embedded with interface technique. The two-step damage identification is verified by a finite element model of Xinxingtang Highway Bridge and a laboratory beam model based on polyvi- nylidens fluoride （PVDF）. The bridge model was constructed with 57 girder segments, and simulated with 58 meas- urement points. The damaged segments were located accurately by GRC index regardless of damage extents and noise levels. With stiffness reduction factors of detected segments as variables, the GA program evolved for 150 generations in 6 h and identified the damage extent with the maximum errors of 1% and 3% corresponding to the noise to signal ratios of 0 and 5%, respectively. In contrast, the common GA-based method without using GRC index evolved for 600 generations in 24 h, but failed to obtain satisfactory results. In the laboratory test, PVDF patches were used as dynamic strain sensors, and the damage locations were identified due to the fact that GRC indexes of points near damaged ele- ments were smaller than 0.6 while those of others were larger than 0.6. The GA-based damage quantification was also consistent with the value of crack depth in the beam model.