Transport Map sampling with PGD model reduction for fast dynamical Bayesian data assimilation

The motivation of this work is to address real-time sequential inference of parameters with a full Bayesian formulation. First, the proper generalized decomposition (PGD) is used to reduce the computational evaluation of the posterior density in the online phase. Second, Transport Map sampling is used to build a deterministic coupling between a reference measure and the posterior measure. The determination of the transport maps involves the solution of a minimization problem. As the PGD model is quasi-analytical and under a variable separation form, the use of gradient and Hessian information speeds up the minimization algorithm. Eventually, uncertainty quantification on outputs of interest of the model can be easily performed due to the global feature of the PGD solution over all coordinate domains. Numerical examples highlight the performance of the method.     

高效乙烯聚合催化剂的制备——I．催化剂的制备规律

研究了新型含镁络合物的结构、性质及其用于制备高效乙烯聚合催化剂的制备规律。结果表明：将镁、有机醇以及不与镁发生“格式反应”的卤代烷相互作用可迅速发生反应形成新型含镁络合物溶液。该络合物在溶液中可析出结晶。具有特定的化学结构．可以均匀承载于多孔硅胶的孔隙中，负载TiCl4后制备出形态良好的球形高效乙烯集合催化剂。     

Reduced order modeling & controller design for mass transfer in a grain storage system

This paper considers the problem of simulating the humidity distributions of a grain storage system. The distributions are described by partial differential equations (PDE). It is quite difficult to obtain the humidity profiles from the PDE model. Hence, a discretization method is applied to obtain an equivalent ordinary differential equation model. However, after applying the discretization technique, the cost of solving the system increases as the size increases to a few thousands. It may be noted that after discretization, the degree of freedom of the system remain the same while the order increases. The large dynamic model is reduced using a proper orthogonal decomposition based technique and an equivalent model but of much reduced size is obtained. A controller based on optimal control theory is designed to obtain an input such that the output humidity reaches a desired profile and also its stability is analyzed. Numerical results are presented to show the validity of the reduced model and possible further extensions are identified.     

大跨度钢网架结构屋面风压预测

风洞试验中,由于结构表面所布置的测点个数远远少于结构节点的个数,因而需要对未布置测点的屋架节点处的风压进行预测．本文采用本征正交分解法利用部分模态对部分测点风压进行了重组,并采用双线性插值和加权的本征正交分解法这两种方法对某海洋馆的屋面风压进行了预测,并对预测结果进行了比较和分析．     

在不完全观察的漂移过程中达到一个目标的最大化概率

对一个具有原始财产X0（〈1）的投资者,当股市的涨落没有被直接观察,而仅仅是用计算的方法建立平均返回扩散模型时,他为了实现一个目标XT=1,如何达到最大的概率？在采用鞅方法的同时,以一个推广的Cameron—Martin公式就能如财富过程一样直接计算价值过程,从而采用Martin公式,可以确定动态的最优配制。     

基于正交分解技术的汽车非定常流场分析及重构

为有效开展汽车气动优化设计和主动流动控制技术,采用正交分解(Proper Orthogonal Decomposition,POD)技术分析汽车非定常流场中POD模态的物理意义,并利用重构技术建立汽车外流场瞬态简化模型.大涡模拟(Large Eddy Simulation,LES)仿真结果与实验结果的对比表明该仿真方法有效.POD分析结果表明:低阶模态主要对应分离泡的附着流动,包含流场中较高的能量;其他各个高阶模态含能相对较低,表征小尺度涡的作用.当保留前15阶POD模态时,通过重构技术可获得70%的原流场能量,得到的降阶模型可以较好地捕捉流动的特征.     

河道生态护坡技术在宜春市城市防洪工程中的运用

根据国内有关城市河道治理的先进经验,采用了生态护坡技术在江西省宜春市城市防洪工程中的应用,取得显著效果.     

论宜春中心城区备用水源建设

江西省宜春中心城区人口增长速度加快,而现状供水规模偏小,且水源单一,一旦遭遇特枯水年份、连续干旱年或袁河发生重大污染事故,中心城区将面临严重的供水危机,急需寻找第二水源点,加快备用水源的建设.     

Vademecum‐based GFEM (V‐GFEM): optimal enrichment for transient problems

This paper proposes a generalized finite element method based on the use of parametric solutions as enrichment functions. These parametric solutions are precomputed off‐line and stored in memory in the form of a computational vademecum so that they can be used on‐line with negligible cost. This renders a more efficient computational method than traditional finite element methods at performing simulations of processes. One key issue of the proposed method is the efficient computation of the parametric enrichments. These are computed and efficiently stored in memory by employing proper generalized decompositions. Although the presented method can be broadly applied, it is particularly well suited in manufacturing processes involving localized physics that depend on many parameters, such as welding. After introducing the vademecum‐generalized finite element method formulation, we present some numerical examples related to the simulation of thermal models encountered in welding processes. Copyright © 2016 John Wiley & Sons, Ltd.     

依赖于真子集的频繁邻近类别集挖掘

针对现有频繁邻近类别集挖掘算法存在重复计算和冗余邻近类别集的问题,提出一种依赖于真子集的频繁邻近类别集挖掘算法,适合在海量数据中挖掘空间对象的频繁邻近类别集。该算法用析构法建立邻近类别集数据库,用产生邻近类别集真子集的方法计算支持数,实现一次扫描数据库提取频繁邻近类别集。算法无需产生候选频繁邻近类别集,且计算支持数时无需重复扫描,从而达到提高挖掘效率的目的。实验结果表明,在海量空间数据中挖掘频繁邻近类别集时,该算法比现有算法更快速有效。