提高前馈神经网络学习效率的学习算法探讨

简要分析了最速下降法（Ｓｔｅｅｐｅｓｔ Ｄｅｓｃｅｎｔ Ａｌｇｏｒｉｔｈｍ ,即ＳＤＡ 法）和正交校正共轭梯度法（ＣＧＭ－ＯＣ法）的优缺点,提出了一种进行多层前馈神经网络学习的新算法,即ＳＤ－ＣＧＭ－ＯＣ算法．该算法结合最速下降法与正交校正共轭梯度法的特点,在文中所述实际问题构造模型的基础上,论证了ＳＤ－ＣＧＭ－ＯＣ算法比传统的ＢＰ算法具有更高的学习效率和二次收敛率．实验结果验证了该学习算法的有效性．     

用多重网格技术加速CELS算法的收敛

求解不可压流体耦合方程的直接解法（ＣＥＬＳ算法）,在许多计算问题中体现了其优 越性。多重网格法在加速数值计算的收敛速度方面也体现出了其有效性。本文把ＣＥＬＳ算法 在多重网格下实施,使用原始变量法,对圆管突扩通道和驱动方腔二维层流流动进行了计算, 并与Ｋｏｓｅｆｆ等人的实验数据进行了比较。结果表明,ＣＥＬＳ算法对所考核的算例是很有效的, 多重网格技术对ＣＥＬＳ算法收敛进程的加这也是显著的。     

分子和表面吸附系统振动频率的理论计算研究

用ＡＳＥＤ－ＭＯ方法计算分子及表面吸附系统在平衡点附近的总能量．用二次多项式、三次多项式和两种形式Ｍｏｒｓｅ势能展开式模拟能量变化以求出力常数，从而得到振动频率．计算结果表明：在势能展开式中引入非对称项是振动频率结果得到较大改善的关键．本文进一步讨论了用Ｍｏｒｓｅ势、ＡＳＥＤ－ＭＯ方法计算的Ｃ２Ｈ２－Ｓｉ（１００）、Ｃ２Ｈ４－Ｓｉ（１００）吸附系统的振动频率，计算结果与实验结果符合得很好，从而支持了ＨＲＥＥＬＳ提出的上述分子在Ｓｉ（１００）－２×１表面的吸附结构．     

公开密钥密码体制RSA算法的一种实现

通过对ＲＳＡ算法的分析和对其几种实现方法的研究比较，针对该算法巨大的计算量开锁问题，对生成密钥对的全过程采用了小素数集测试、优化的Ｍｉｎｅｒ－Ｒａｂｉｔ测试、Ｓｔｅｉｎ算法、求解同余方程等一系列方法，给出了该算法运行速度较高的一种实现，在ＡＭＤ２６６ ＣＰＵ、６４ＭＢ内存、Ｗｉｎｄｏｗｓ９８操作系统的测试环境下，选取２加密强度（即ＲＳＡ算法的模长）为１０２４ｂｉｔ时，生成ＲＳＡ算法密钥对的运行时间     

激波绕尖劈障碍物运动粘性流场的数值模拟研究

本文以数值方法研究了运动激波绕射的非定常粘性流的发展过程．计算中采用二阶ＴＶＤ差分格式求解二维非定常薄层Ｎａｖｉｏｒ－Ｓｔｏｋｅｓ方程，得到了清晰的流场结构图谱，分析了产生局部粘性分离的现象．研究结果表明，本文的方法可以很好地描述激波绕射的复杂流动，对流场间断有较高的捕获能力，同时表明在激波障碍物干扰的绕射场研究中局部粘性效应作用是不可忽视的重要因素     

_2(R,S)中的 Parsimonious-矩阵

设Ａ２（Ｒ,Ｓ）是所有具有指定行和向量Ｒ与列和向量Ｓ的（０,１,２）－矩阵组成的集合．Ａ２（Ｒ,Ｓ）中正元素个数最少的矩阵称为是Ｐａｒｓｉｍｏｎｉｏｕｓ－矩阵．本文主要研究Ａ２（Ｒ,Ｓ）中Ｐａｒｓｉｍｏｎｉｏｕｓ－矩阵的性质,并给出一种找出Ｐａｒｓｉｍｏｎｉｏｕｓ－矩阵的简捷算法．     

钢纤维混凝土圆柱壳稳定性能的试验研究

本文设计创作了一个钢纤维混凝土圆柱薄壳，采用一套特制的杠杆加载体系对其稳定性能进行了测试。利用自行研制的非线性计算程序ＮＡＰＳ－ｌ、ＮＡＰＳ－２（ＮｏｎｌｉｎｅａｒＡｎａｌｙｓｉｓＰｒｏｇｒａｍｆｏｒＳｈｅｌｌｓ）对该薄壳结构进行了几何非线性和一般非线性全过程分析，并将试验结果与理论计算结果进行了分析比较。     

热处理共聚醚酯的X光衍射研究

采用ＳＡＸＳ和ＷＡＸＳ技术，对不同温度热处理的ＰＢＴ－ＰＥＧ无规多嵌段共聚醚酯的相变化进行研究．结果表明，随着热处理温度的提高，结晶区体积增大，且主要是侧向增长，同时无定形区的体积也增加，两相间过渡层变薄，趋于形成理想的两相结构．     

离心式压缩机内三维粘性流动计算

采用时间相关的可压缩平均Ｎ－Ｓ方程组及Ｋ方程湍流模型，应用任意拉格朗－欧拉法（ＡＬＥ）进行离散，对离心压缩机叶轮内部的三维粘性流场进行了数值模拟，ＡＬＥ法采用任意六面体网络单元，网格可按规定速度运动，因此适用于具有复杂几何边界和流场区域不断变化的流动问题，以Ｅｃｋａｒｄｔ叶轮为例进行了数值计算，计算结果与实验结果吻合良好。     

一种新的SRA图像Speckle滤波算法

详细分析了合成孔径雷达图像中相干斑的统计特性，提出了一种具有窗口自选功能的自适应相干斑滤波算法，并将该算法用于Ｓｅａｓａｔ－ＡＳＡＲ图像的滤波处理；并将滤波结果与Ｌｅｅ算法Ｏｄｄｙ算法作了对比试验，其结果表明滤波性能地Ｌｅｅ算法和Ｏｄｄｙ算法。     

Numerical Simulation of Twin-Screw Extrusion with Wall Slip

Wall slip boundary condition is first introduced into twin-screw extrusion with the Navier slip law. Three-dimensional isothermal flow in the twin-screw extruder is simulated by using the finite element package POLYFLOW. Profiles of velocity contours in the screw channel and shear rate distributions in the intermeshing region are presented for different slip coefficients. Curves of axial pressure difference, average shear rate and dispersive mixing index vs. the slip coefficient are plotted and discussed. Comparisons are also made between the wall slip conditions and the non-slip condition. The simulation results indicate that, as the level of wall slip decreases, the axial pressure difference rises, the shear effect is intensified and the axial mixing is also enhanced. All these flow characteristics seem to level off with the increase of the slip coefficient. However, because of the inherent limitation of the Navier slip law, use of an overestimated slip coefficient would predict an over-sticky state between the screw surface and the polymer melt.     

Numerical simulation of electroosmotic flow in hydrophobic microchannels

Electroosmotic flow (EOF) is a promising way for driving and mixing fluids in microfluidics. For the parallel-plate microchannel with the hydrophobic surface, this paper solved the governing equations using the finite element method (FEM), and the effects of microchannel height, electric strength and ionic concentration on EOF were thus investigated. The simulation indicates that the transient characteristics of EOF are similar in hydrophobic and hydrophilic microchannels, the steady time of EOF is proportional to the square of microchannel height, and the scale is microsecond. EOF velocity is proportional to the electric strength and independent of the channel height, and decreases slowly with the ionic concentration, which is lower than that in hydrophilic microchannel due to the presence of slip length in hydrophobic microchannel. The results can provide valuable insights into the optimal design of microchannel surfaces to achieve accurate EOF control in hydrophobic microchannel. Supported by the National Natural Science Foundation of China (Grant No. 50730007)     

圆管分层层流的流动规律

从流体力学中N－S方程出发，建立了圆管两相分层层流流动的数学模型，通过两相分层层流的相互作用的流体力学分析，提出了两相流动界面的耦合条件，从而得到了一组完整描述两相圆管分层流的数学方程，再通过数值求解分层流的数学方程，得到了对工程实际应用有重要价值的流动规律，应用圆管两相分层层流的流动规律，可以实现高粘流体的高效率输送，其节能效果十分明显。     

Fluid Flow Behavior of Liquid in Cylindrical Vessels Stirred by One or Two Air Jets

Based on the two-phase model (Eulerian-Eulerian model), the three dimensional fluid flow in water and that liquid steel systems stirred by one or two multiple gas jets are simulated. In the Eulerian-Eulerian two-phase model, the gas and the liquid phase are considered to be two different continuous fluids interacting with each other through the finite inter-phase areas. The exchange between the phases is represented by source terms in conversation equations. Turbulence is assumed to be a property of the liquid phase. A new turbulence modification - model is introduced to consider the bubbles movement contribution to and . The dispersion of phases due to turbulence is represented by introducing a diffusion term in mass conservation equation. The mathematical simulation agrees well with the experiment results. The study results indicate that the distance of two nozzles has big effect on fluid flow behavior in the vessel. Using two gas injection nozzles at the half radii of one diameter of the bottom generates a much better mixing than with one nozzle under the condition of the same total gas flow rate.     

Slip flow and variable properties of viscoelastic fluid past a stretching surface embedded in a porous medium with heat generation

This study examines theoretically and computationally the non-Newtonian boundary layer flow and heat transfer for a viscoelastic fluid over a stretching continuous sheet embedded in a porous medium with variable fluid properties, slip velocity, and internal heat generation/absorption. The flow in boundary layer is considered to be generated solely by the stretching of the sheet adjacent to porous medium with boundary wall slip condition. Highly nonlinear momentum and thermal boundary layer equations governing the flow and heat transfer are reduced to set of nonlinear ordinary differential equations by appropriate transformation. The resulting ODEs are successfully solved numerically with the help of shooting method. Graphical results are shown for non-dimensional velocities and temperature. The effects of heat generation/absorption parameter, the porous parameter, the viscoelastic parameter, velocity slip parameter, variable thermal conductivity and the Prandtl number on the flow and temperature profiles are presented. Moreover, the local skin-friction coefficient and Nusselt number are presented. Comparison of numerical results is made with the earlier published results under limiting cases.     

Thermodynamic properties and mixing thermodynamic parameters of two-phase metallic melts

Based on the calculating model of metallic melts involving eutectic, the calculating equations of mixing thermodynamic parameters for two phase metallic melts have been formulated in the light of those equations of homogeneous solutions. Irrespective as to whether the activity deviation relative to Raoultian behavior is positive or negative, or the deviation is symmetrical or unsymmetrical, the evaluated results not only agree well with experimental values, but also strictly obey the mass action law. This testifies that these equations can authentically reflect the structural reality and mixing thermodynamic characteristics of two-phase metallic melts. The calculating equations of mixing thermodynamic parameters for the model of two phase metallic melts offer two practical criteria (activity and mixing thermodynamic parameters) and one theoretical criterion (the mass action law).     

水平井早期试井公式

水平井早期流动是指水平井从开始生产起到压力降传播波尚达到储藏边界的阶段。在已有献中早期阶段被视为径向流，早期结束时间也存在各种计算公式，不同公式的结果相差甚大，并且皆缺少理论依据。根据早期时刻的等势面是以水平井两端点焦点的旋转椭球面，并应用质量守恒定律导出早期试井公式和早期结束时间的算法，理论和实例皆表明所推导的公式更能描述早期渗流规律，计算结果更符合实际情况。     

沙尘天气模拟系统混合段长度计算

应用气固两相流理论计算了沙尘天气模拟系统颗粒自由悬浮速度和混合段长度,利用FLUENT对该系统气固两相流场进行了数值模拟,其中连续相、颗粒相分别采用k-ε湍流模型和离散相模型,相问作用采用随机跟踪模型.结果表明,连续相流场分布规律与理论设计吻合较好,颗粒相浓度场分布可达到试验设计要求.     

无粘结预应力钢板夹心混凝土组合板滑移性能分析

目的钢板与混凝土界面相对滑移是影响组合板承载力和刚度的主要因素,明确界面相对滑移分布规律及影响因素.方法利用弹性理论,建立考虑预应力钢筋内力增量和滑移刚度影响的界面相对滑移微分方程,给出对称集中荷载作用下的滑移计算公式.结果组合板最大滑移发生在组合板端部,并沿组合板纵向向跨中逐渐减小,跨中滑移为零,组合板的滑移随着连接刚度增加、含钢率增加及混凝土强度等级提高而减小.结论连接刚度是影响预应力组合板界面滑移的主要因素,混凝土强度等级、含钢率影响相对较小,预应力钢筋内力增量对滑移影响很小,可以忽略其影响.     

悬浮烘干机冷态流场的数值计算

悬浮烘干机是一种新型的烘干设备,目前已在一些工厂投入使用。本文所研究的是该烘干机内的冷态流场,按照 Boussinesp 假说和混合长度理论,作者在对该流场进行了四项合理性假定的基础上,将雷诺方程简化为一个边值问题和一个初值问题。对此边、初值问题进行求解,得到了实测结果相吻合的理论计算结果。