一种基于PETSc的热传导方程大规模并行求解策略

提出了一种大规模热传导方程并行求解的策略,采用了分布式内存和压缩矩阵技术解决超大规模稀疏矩阵的存储及其计算,整合了多种Krylov子空间方法和预条件子技术来并行求解大规模线性方程组,基于面向对象设计实现了具体应用与算法的低耦合.在Linux机群系统上进行了性能测试,程序具有良好的加速比和计算性能.     

三维热传导方程的Krylov子空间方法并行分析

热传导方程在地下水流动数值模拟、油藏数值模拟等工程计算中有着广泛应用,其并行实现是加速问题求解速度、提高问题求解规模的重要手段,因此热传导方程的并行求解具有重要意义。对Krylov子空间方法中的CG和GMRES算法进行并行分析,并对不同的预处理CG算法作了比较。在Linux集群系统上,以三维热传导模型为例进行了数值实验。实验结果表明,CG算法比GMRES算法更适合建立三维热传导模型的并行求解。此外,CG算法与BJACOBI预条件子的整合在求解该热传导模型时,其并行程序具有良好的加速比和效率。因此,采用BJACOBI预处理技术的CG算法是一种较好的求解三维热传导模型的并行方案。     

基于光滑聚集代数多重网格的有限元 并行计算实现方法

基于光滑聚集代数多重网格法实现一种用于结构有限元并行计算的预条件共轭梯度求解方法。对计算区域进行均匀划分,将这些子区域分配给各个进程同时进行单元刚度矩阵的计算,并组合形成分布式存储的整体平衡方程。采用光滑聚集代数多重网格预条件共轭梯度法对整体平衡方程进行并行求解,在天河二号超级计算机上进行数值试验,分析代数多重网格的主要参数对算法性能的影响,测试程序的并行计算性能。试验结果表明该方法具有较好的并行性能和可扩展性,适合于大规模实际应用。     

稀疏近似逆预条件子及其并行计算

文中使用范数极小技术,提出一种构造稀疏矩阵并行近似逆预条件子的方法,所构造的稀疏矩阵近似逆的稀疏结构和数据矩阵的转置矩阵相同,计算量和存储量上,其求解过程易于并行。且并行计算不影响其收敛效果。通过试算表明,该方法对很多问题的求解具有明显的加速效果。文中给出了该方法的并行算法,并提出了一种自适应分配算法来解决负载平衡问题。     

稀疏线性方程组求解中的预处理技术综述

稀疏线性方程组的高效求解是数值计算方向的研究热点之一,其中包括预处理技术的研究。本文从技术分类的角度,总结了稀疏线性方程组求解中的预处理技术。首先,介绍了填充元缩减策略,旨在减少求解过程中存储量的同时,仍能保持矩阵的稀疏结构;其次,介绍了不同结构系数矩阵的多种匹配技术,旨在获得矩阵的对角优势性;最后,介绍了具有天然并行性的因子分解近似逆预条件子构造方法和不完全分解预条件中的并行求解技术等。     

基于数值计算模型的MMC半桥型子模块仿真方法

模块化多电平换流器（MMC）的各半桥子模块均由两个开关组（1个IGBT和1个反并联的二极管）构成。针对MMC在包含的子模块规模较大的情况下,对采用电路模型分割法对MMC进行分割后的子模块进行仿真求解时仍然会占用较多资源,效率不高的问题,提出了基于数值计算模型的MMC半桥子模块仿真验证方法。首先通过分析三相MMC及其半桥子模块（HBSM）的工作机制,把半桥型子模块中的两个开关组等效为在高、低阻态不断切换的等效电阻并给出了其等效电路。然后针对电容支路的离散化问题,根据梯形积分法推导了MMC半桥型子模块的数值求解公式,给出了数值计算电路模型。最后基于MATLAB仿真平台建立了基于数值计算模型的半桥子模块仿真验证模型,通过与详细模型子模块的仿真波形对比分析,结果表明了所建立的子模块数值计算模型是可行的。     

重叠组合法的芯片级三维寄生电容提取及其并行实现

采用双向区域重叠组合法，基于三维层次式块边界元法实现了芯片级的互连电容提取．该方法将芯片切分为大量小规模区域。用全局场求解器计算各子区域电容矩阵，可方便地组合出整个芯片的电容矩阵；同时分析了其计算量和精度，并进行了并行计算实验．对实际版图结构的数值实验验证了有关分析结论，表明该方法高效、可靠、并行性能好．     

一种针对大波数Helmholtz方程的高性能并行预条件迭代求解算法

针对传统串行迭代法求解大波数Helmholtz方程存在效率低下且受限于单机内存的问题,提出了一种基于消息传递接口(Message Passing Interface,MPI) 的并行预条件迭代法。该算法利用复移位拉普拉斯算子对Helmholtz方程进行预条件处理,联合稳定双共轭梯度法和基于矩阵的多重网格法来求解预条件方程离散后的大规模线性系统,在Linux集群系统上基于 MPI环境实现了求解算法的并行计算,重点解决了多重网格的并行划分、信息传递和多重网格组件的构建问题。数值实验表明,对于大波数问题,提出的算法具有良好的并行加速比,相较于串行算法极大地提高了计算效率。     

基于曙光并行机的超大规模非线性方程组并行算法研究

该文讨论了一类求解大规模非线性方程组算法的并行性能及其在曙光并行机上的实现过程，与传统的算法不同之处是用一个块对角矩阵作为迭代矩阵，且该矩阵可由一个仅包含向量内积和矩阵与向量乘积的递推关系简便计算得到，在对算法进行描述之后，分析了算法的并行加速比和存储需求，讨论了算法在基于消息传递的MPI并行环境下的实现流程，数值计算表明理论分析与数值结果相比，算法在分布式并行环境下具有有较好的并行主攻较低的存储要求，可适用于大规模科学与工程的高性能计算。     

一种用计算域分解的等几何分析并行化方法

提出一种按照计算域分解的并行化方法来构建等几何分析的刚度矩阵和右侧向量.将计算域分解成为若干个不相交的子区域,然后为每个区域分配一个处理器,所有处理器并行进行子区域上面的计算,所有处理器完成子区域的计算以后,使用一个快速的归并算法完成线性系统的装配.实验表明,本文提出的方法在8核的机器上可以达到6.46的加速比,能够在4秒左右的时间计算680万个矩阵元素个数.使用Intel MKL稀疏求解器来求解线性系统,本文的等几何分析求解器能够在大约10秒的时间内求解52万的自由度,本文的方法比ISOGAT速度要快上万倍.     

Moment stability for linear systems with a random parametric excitation

Moment stability for linear systems with a nonwhite parametric noise is considered. A method of reduction of the study of this stability to the study of stability for large-scale matrices is proposed. Mean square stability diagrams for random harmonic oscillator are presented.     

基于广义模型的复杂电力系统的故障诊断

讨论了广义大系统故障诊断的一种新方法。首先,给出了复杂电力系统不对称故障的定义,以及传统的分析方法。接着,对所给的广义大系统,构造它的一个全阶广义状态观测器。考虑广义系统的3种可能故障形式,残差由交联项部分和系统的故障两部分组成。为了排除交联项对故障的干扰,实现交联项的解耦,分两种情况讨论残差,并分别设计加权矩阵,给出故障诊断的方法。所给的故障诊断方法在实际应用中切实可行,且能提高故障诊断效率。     

Transient analysis of flexible multi-body systems. Part I: Dynamics of flexible bodies

The formulation for the dynamic analysis of undamped linear structural systems using the finite element method results in two element matrices; the mass and stiffness matrices, that describe the element inertia and stiffness properties. However, these matrices are not sufficient to describe the dynamics of structures that undergo large rigid-body motion. Other element matrices, in addition to the mass and stiffness matrices, are required to account for the inertia coupling between gross motion and elastic deformation. These matrices are time-invariant and can be generated and assembled in the same manner as the mass and stiffness matrices are assembled in linear structural dynamics. An inherent relation between these matrices and the deformable body mean axes exists. This paper is the first of two parts. It presents the two-dimensional and three-dimensional formulation of the system equations of motion of inertia-variant flexible bodies. In particular, Euler parameters are employed to describe the rotations of the body reference in the spatial analysis. In Part II [13], this formulation is applied to the impact analysis of a large-scale constrained flexible aircraft which are modeled as a multi-body system consisting of interconnected rigid and flexible components.     

Logarithmic barriers for sparse matrix cones

Algorithms are presented for evaluating gradients and Hessians of logarithmic barrier functions for two types of convex cones: the cone of positive semidefinite matrices with a given sparsity pattern and its dual cone, the cone of sparse matrices with the same pattern that have a positive semidefinite completion. Efficient large-scale algorithms for evaluating these barriers and their derivatives are important in interior-point methods for nonsymmetric conic formulations of sparse semidefinite programs. The algorithms are based on the multifrontal method for sparse Cholesky factorization.     

swSpAMM: optimizing large-scale sparse approximate matrix multiplication on Sunway Taihulight

Although matrix multiplication plays an essential role in a wide range of applications, previous works only focus on optimizing dense or sparse matrix multiplications. The Sparse Approximate Matrix Multiply (SpAMM) is an algorithm to accelerate the multiplication of decay matrices, the sparsity of which is between dense and sparse matrices. In addition, large-scale decay matrix multiplication is performed in scientific applications to solve cutting-edge problems. To optimize large-scale decay matrix multiplication using SpAMM on supercomputers such as Sunway Taihulight, we present swSpAMM, an optimized SpAMM algorithm by adapting the computation characteristics to the architecture features of Sunway Taihulight.Specifically, we propose both intra-node and inter-node optimizations to accelerate swSpAMM for large-scale execution. For intra-node optimizations, we explore algorithm parallelization and block-major data layout that are tailored to better utilize the architecture advantage of Sunway processor. For inter-node optimizations, we propose a matrix organization strategy for better distributing sub-matrices across nodes and a dynamic scheduling strategy for improving load balance across nodes. We compare swSpAMM with the existing GEMM library on a single node as well as large-scale matrix multiplication methods on multiple nodes. The experiment results show that swSpAMM achieves a speedup up to 14.5× and 2.2× when compared to xMath library on a single node and 2D GEMM method on multiple nodes, respectively.     

PARAFAC algorithms for large-scale problems

Parallel factor analysis (PARAFAC) is a tensor (multiway array) factorization method which allows to find hidden factors (component matrices) from a multidimensional data. Most of the existing algorithms for the PARAFAC, especially the alternating least squares (ALS) algorithm need to compute Khatri-Rao products of tall factors and multiplication of large matrices, and due to this require high computational cost and large memory and are not suitable for very large-scale-problems. Hence, PARAFAC for large-scale data tensors is still a challenging problem. In this paper, we propose a new approach based on a modified ALS algorithm which computes Hadamard products, instead Khatri-Rao products, and employs relatively small matrices. The new algorithms are able to process extremely large-scale tensors with billions of entries. Extensive experiments confirm the validity and high performance of the developed algorithm in comparison with other well-known algorithms.     

Design of state feedback for large-scale multivariable systems

This note presents a general method which reduces the computational requirements in the state feedback design of large-scale multivariable systems. The given system is first transformed into a general block canonical form by using simple equivalent transformation. The state feedback problem is then reformulated in terms of a Sylvester equation. Finally, the transformed system matrices along with certain assumed block forms for unknown matrices enable the Sylvester equation to be decomposed and solved effectively.     

数值界不确定性关联大系统分散鲁棒H∞控制

针对一类状态阵,控制输入阵及关联阵中存在数值界不确定性的关联大系统,研究其分散鲁棒H∞状态反馈和输出反馈控制器设计问题.基于有界实引理,推导出了其存在分散鲁棒H∞控制器的充分条件,即一组矩阵不等式有解.利用Schur补引理,通过固定不同变量,提出了一种构建分散控制器的同伦迭代线性矩阵不等式方法.所获得的控制器使闭环大系统鲁棒稳定,并且达到给定的H∞性能指标.最后用数值例子说明了所提的设计方法的有效性.     

线性非定常系统的稳定性问题

大系统的理论与应用近十余年来有了相当大的发展,本文研究了这类系统的稳定性问题。首先对非定常线性系统的稳定性给出了一个简单的几何判据,然后建立起大系统的稳定性判据。最后考虑了大系统的结构,从而建立了简化的稳定性判据。     

关联大系统的分散H∞/LTR控制

讨论关联大系统分散回路传递再生的设计问题.利用矩阵的奇异值分解技术,提出了对 关联项进行块对角化处理的一种新方法.基于H∞理论,将多变量系统的回路传递再生方法 推广于分散控制关联大系统,并避免了在所有可能的分散观测器增益构成的集合上,直接计算 再生矩阵的H∞范数下确界的困难.