最优化算法的收敛准则

收敛准则是最优化算法的重要组成部分，其选择得好与坏将直接影响到算法的成功与否以及收敛得快与慢。现有常用的收敛准则基本上是建立在前后迭代点的逼近和它们相应函数值的逼近是否达到一定的精度要求以及迭代点处函数梯度是否接近于零的基础上的。它们各自有自己的适用范围。但它们的共同特点是对迭代终止点的性质不能做出判断。本文在总结和分析现有算法收敛准则的基础上，借助于正定矩阵、一维优化方法中对分法和黄金分割法，提出了新的算法收敛准则。算例结果表明，这些收敛准则是有效实用的。     

关于一类电-力振动模型的建模与分析

从电学、力学的基本原理出发,通过数学方法建立和分析了一种电-力振动模型。这是一种较复杂的电力混合作用的线性振动系统,由模型的特殊结构(类扬声器结构)作者确立了两种磁场,即感生磁场和外磁场两者正交独立,并规定了电学、力学两种不同物理量的坐标取向关系。该模型需要求得三阶正系数常微分方程的收敛解,再求得包含暂态、稳态项的完整解。另外,文章从能量和做功的角度,通过对电压电流间的相位差分析,对所建模型的正确性作了论证,同时也为这类建模引荐了一种论证手段。     

RENEWAL OF BASIC LAWS AND PRINCIPLES FOR POLAR CONTINUUM THEORIES (Ⅲ) —NOETHER’S THEOREM

The existing various couple stress theories have been carefully restudied.The purpose is to propose a coupled Noether’s theorem and to reestablish rather complete conservation laws and balance equations for couple stress elastodynamics. The new concrete forms of various conservation laws of couple stress elasticity are derived. The precise nature of these conservation laws which result from the given invariance requirements are established. Various special cases are reduced and the results of micropolar continua may be naturally transited from the results presented in this paper.     

Weighted Cebysev-Ostrowski type inequalities

In account of the famous Cebysev inequality, a rich theory has appeared in the literature. We establish some new weighted Cebysev type integral inequalities. Our proofs are of independent interest and provide new estimates on these types of inequalities.     

基于可靠度的结构优化的序列近似规划算法

基于可靠度的优化的最直观解法是把可靠度和优化的各自算法搭配一起形成嵌套两层次迭代。为改善其收敛性提高计算效率,人们提出了功能测度法、半无限规划法、单层次算法等多种改进方法。本文对传统结构优化界的经典序列近似规划法改造并扩展应用于求解基于可靠度的结构优化问题,构造该问题的序列近似规划模型和求解过程;其核心思想是在每个近似规划子问题中采用近似可靠度指标对设计变量的线性近似,在优化迭代过程中同步更新设计变量和随机空间中的近似验算点坐标,以达到可靠度分析和优化迭代同步收敛的目标。为了算法的实施,还推导出近似可靠度指标的半解析灵敏度计算公式,编制了程序,最终实现与通用软件的连接。论文用算例证实算法的有效性。     

INTERACTION OF ELECTRIC CHARGES IN A PIEZOELECTRIC WITH RIGID EXTERNAL CRACKS

IntroductionRecently much attention has been paid to Green’s functions[1 -7]for anisotropic mediabecause of their important applications in boundary element methods used widely inengineering analysis. Generally speaking, it is relatively easy to obtain G…     

部分输入未知条件下结构动力复合反演的分解算法

在充分利用部分输入已确知而部分输入未知的激励特性的基础上，提出了结构动力复合反演的分解算法，该算法从源头上消除了迭代过程中参数识别与荷载反演的相互影响，降低了问题的计算规模。对于线性参数系统，该算法不经过任何迭代计算即可一次性完成结构参数识别及荷载反演。将其与松弛法结合，可解决非线性参数系统的识别问题，与文献[4]的方法比较，其收敛速度有显著提高。     

Branching fractions and direct C P asymmetries of ■_s →K~0h~+h′~-(h(′)=K,π) decays

Motivated by the recent LHCb collaboration measurements of charmless three-body decays of ■s meson, we calculate the branching fractions of ■ s →K0π+π-, ■s→K0K+K-, ■s→K0π+K- and ■s→K0K+π- decay modes using the factorization approach.Both the resonant and nonresonant contributions are studied in detail. For the decays ■s→K0π+π- and ■s→K0K+K-, our results agree well with experimental data, and the former is dominated by the K*, while the latter one is dominated by the nonresonant contribution. Considering the flavor S U(3) symmetry violation, the sum of branching fractions of ■s→K0π+K- and ■s→K0K+π-could accommodate the data well too. It should be noted that both branching fractions are sensitive to the scalar density Kπ|ˉsq|0.Furthermore, the resonant contributions are dominated by the scalar K*0(1430). We hope that these branching fractions could be measured individually in the experiments so as to test the factorization approach and the flavor S U(3) asymmetry. Moreover, the direct C P asymmetries of these decays are also investigated, which could be measured in the running LHCb experiment and Super-b factory in the future.     

三步混合型合成隐迭代序列的强收敛定理

在Banach空间中,引入有限族渐近非扩张自映射和渐近非扩张非自映射的新的三步合成隐迭代序列.并证明该迭代序列的强收敛定理.     

具有变时滞和马尔可夫切换的随机递归神经网络的弱收敛(英文)

本文研究了具有变时滞和马尔可夫切换的随机递归神经网络的弱收敛,通过运用Lyapunov函数、随机分析技巧和推广了的Halanay不等式,得到了上述模型为弱收敛的充分性条件,并且我们揭示了对上述递归神经网络模型所确定的segment过程的转移概率的极限分布是此模型的解过程的唯一的遍历不变概率测度.此外,我们还给出了例子和数值模拟来说明我们结论的正确性.