FAST ALGORITHMS FOR DETERMINING POSITIVE DEFINITENESS OF LARGE SCALE MATRICES IN SOLID STRUCTURE ANALYSIS

对于结构稳定性分析中超大规模矩阵正定性判定,必须采用并行计算方法,传统方法如计算特征值、主子式行列式及LDLT等直接方法难以实现.本文提出了一些可适用于并行的迭代判定算法.借鉴力学系统中能量下降的思想,发展了一种判定矩阵正定性的新思路,即将矩阵的正定性判定问题转化为一个优化问题,并基于优化算法来判定矩阵的正定性.提出了基于最速下降法和共轭梯度法来进行矩阵正定性判定的算法.然后考虑到力学系统刚度矩阵的稀疏性和结构刚失稳状态的弱非正定性,提出可以先截超平面后解方程求驻值点的方法来判定弱非正定矩阵的正定性.为了保证对强非正定矩阵判定的准确性,本文提出可以高效混杂使用截平面法和共轭梯度法.数值实验结果表明,本文提出的算法具有准确性和高效性.对于强非正定矩阵而言,共轭梯度算法更加高效;而对于弱非正定矩阵,则是截平面法和混杂算法更加高效.这些算法都容易在机群上实现并行计算,能够快速判定大规模矩阵的正定性.     

大规模并行结构动力分析分层计算方法

多核分布式存储超级计算机的兴起为大规模并行结构动力分析提供了强有力的计算工具。根据多核分布式计算环境的特点,提出了一种大规模并行结构动力分析分层计算方法。该方法在传统隐式动力分析的区域分解法的基础上,利用两级分区和两次缩聚策略进行求解。不但通过进一步缩减求解问题规模有效提高了界面方程的收敛速度,而且通过三层并行计算有效提高了通信效率。该方法并不对有限元模型引入近似,属于精确的动力子区域分层计算方法。典型数值算例表明,该方法计算精度与商业软件ANSYS完全法求解精度相当;同传统区域分解法相比,该方法能够获得较高的并行计算性能。     

Connection machine simulation of ultrasonic wave propagation in materials. II: The two-dimensional case

A local interaction simulation approach (LISA) for the wave propagation in inhomogeneous 2D media is presented. The method is designed to take full advantage of massively parallel computing, such as provided by the Connection Machine. Crosspoints at the intersection of orthogonal interfaces separating media of different physical properties are treated in the framework of a sharp interface model. A comparison with finite difference techniques shows that the proposed method avoids the ambiguities due to the smoothing of the physical quantities, which is necessary in order to transform differential equations into finite difference equations. The smoothing procedure may cause severe numerical errors, when the variations of the physical properties across the interfaces are large.

In order to demonstrate the efficiency and reliability of the approach several examples of simulation of pulse propagation in different media are reported.     

A HIERARCHICAL PARALLEL COMPUTING APPROACH FOR STRUCTURAL STATIC FINITE ELEMENT ANALYSIS

根据分布式存储并行计算机的体系结构特点,提出了一种结构静力有限元分层并行计算方法. 该方法在两级分区两次缩聚策略的基础上不仅实现了大量数据的分布式存储,提高了数据的内存访问速率;而且实现了计算过程的三层并行,有效提高了通信效率;此外,它还进一步降低了界面方程的规模,大幅度减少了界面方程的求解时间. 因此,它能够充分利用分布式存储并行计算机的体系结构特点提升大规模并行计算效率. 最后通过典型数值算例验证了该方法的正确性和有效性.     

结构静力有限元分层并行计算方法

根据分布式存储并行计算机的体系结构特点,提出了一种结构静力有限元分层并行计算方法. 该方法在两级分区两次缩聚策略的基础上不仅实现了大量数据的分布式存储,提高了数据的内存访问速率;而且实现了计算过程的三层并行,有效提高了通信效率;此外,它还进一步降低了界面方程的规模,大幅度减少了界面方程的求解时间. 因此,它能够充分利用分布式存储并行计算机的体系结构特点提升大规模并行计算效率. 最后通过典型数值算例验证了该方法的正确性和有效性.      

Floquet–Bloch decomposition for the computation of dispersion of two-dimensional periodic,damped mechanical systems

Floquet–Bloch theorem is widely applied for computing the dispersion properties of periodic structures, and for estimating their wave modes and group velocities. The theorem allows reducing computational costs through modeling of a representative cell, while providing a rigorous and well-posed spectral problem representing wave dispersion in undamped media. Most studies employ the Floquet–Bloch approach for the analysis of undamped systems, or for systems with simple damping models such as viscous or proportional damping. In this paper, an alternative formulation is proposed whereby wave heading and frequency are used to scan the k-space and estimate the dispersion properties. The considered approach lends itself to the analysis of periodic structures with complex damping configurations, resulting for example from active control schemes, the presence of damping materials, or the use of shunted piezoelectric patches. Examples on waveguides with various levels of damping illustrate the performance and the characteristics of the proposed approach, and provide insights into the properties of the obtained eigensolutions.     

A high‐order backward forward sweep interpolating algorithm for semi‐Lagrangian method

Conventional semi‐Lagrangian methods often suffer from poor accuracy and imbalance problems of advected properties because of low‐order interpolation schemes used and/or inability to reduce both dissipation and dispersion errors even with high‐order schemes. In the current work, we propose a fourth‐order semi‐Lagrangian method to solve the advection terms at a computing cost of third‐order interpolation scheme by applying backward and forward interpolations in an alternating sweep manner. The method was demonstrated for solving 1‐D and 2‐D advection problems, and 2‐D and 3‐D lid‐driven cavity flows with a multi‐level V‐cycle multigrid solver. It shows that the proposed method can reduce both dissipation and dispersion errors in all regions, especially near sharp gradients, at a same accuracy as but less computing cost than the typical fourth‐order interpolation because of fewer grids used. The proposed method is also shown able to achieve more accurate results on coarser grids than conventional linear and other high‐order interpolation schemes in the literature. Copyright © 2017 John Wiley & Sons, Ltd.     

基于并行计算和遗传算法的钢-UHPC华夫板组合梁优化设计

为实现钢-超高性能混凝土(UHPC)华夫板组合梁结构快速经济合理的设计,提出了基于并行计算与遗传算法的结构优化设计方法。通过Python建立了并行计算平台,使Abaqus和Python能够执行同步数值模拟和数据处理,以成本最小化为目标,采用遗传算法对钢-UHPC华夫板组合梁进行了优化,验证了所提方法的可行性。结果表明,遗传算法中密集的分析任务可以并行化并分配给不同的计算资源以提高计算效率;使用并行计算可以提高8.6倍的优化效率;并行计算和串行计算的CPU平均使用率分别为82%和18%。本文方法的成功应用可为其他类型结构的优化设计提供参考。     

PARALLEL COMPUTING FOR STATIC RESPONSE ANALYSIS OF STRUCTURES WITH UNCERTAIN-BUT-BOUNDED PARAMETERS

The vertex solution for estimation on the static displacement bounds of structures with uncertain-but-bounded parameters is studied in this paper. For the linear static problem, when there are uncertain interval parameters in the stiffness matrix and the vector of applied forces, the static response may be an interval. Based on the interval operations, the interval solution obtained by the vertex solution is more accurate and more credible than other methods （such as the perturbation method）. However, the vertex solution method by traditional serial computing usually needs large computational efforts, especially for large structures. In order to avoid its disadvantages of large calculation and much runtime, its parallel computing which can be used in large-scale computing is presented in this paper. Two kinds of parallel computing algorithms are proposed based on the vertex solution. The parallel computing will solve many interval problems which cannot be resolved by traditional interval analysis methods.     

A massively parallel GPU‐accelerated model for analysis of fully nonlinear free surface waves

We implement and evaluate a massively parallel and scalable algorithm based on a multigrid preconditioned Defect Correction method for the simulation of fully nonlinear free surface flows. The simulations are based on a potential model that describes wave propagation over uneven bottoms in three space dimensions and is useful for fast analysis and prediction purposes in coastal and offshore engineering. A dedicated numerical model based on the proposed algorithm is executed in parallel by utilizing affordable modern special purpose graphics processing unit (GPU). The model is based on a low‐storage flexible‐order accurate finite difference method that is known to be efficient and scalable on a CPU core (single thread). To achieve parallel performance of the relatively complex numerical model, we investigate a new trend in high‐performance computing where many‐core GPUs are utilized as high‐throughput co‐processors to the CPU. We describe and demonstrate how this approach makes it possible to do fast desktop computations for large nonlinear wave problems in numerical wave tanks (NWTs) with close to 50/100 million total grid points in double/single precision with 4 GB global device memory available. A new code base has been developed in C++ and compute unified device architecture C and is found to improve the runtime more than an order in magnitude in double precision arithmetic for the same accuracy over an existing CPU (single thread) Fortran 90 code when executed on a single modern GPU. These significant improvements are achieved by carefully implementing the algorithm to minimize data‐transfer and take advantage of the massive multi‐threading capability of the GPU device. Copyright © 2011 John Wiley & Sons, Ltd.     

Parallel‐in‐time simulation of the unsteady Navier–Stokes equations for incompressible flow

In this paper, we present an application of a parallel‐in‐time algorithm for the solution of the unsteady Navier–Stokes model equations that are of parabolic–elliptic type. This method is based on the alternated use of a coarse global sequential solver and a fine local parallel one. A standard finite volume/finite differences first‐order approach is used for discretization of the unsteady two‐dimensional Navier–Stokes equations. The Taylor vortex decay problem and the confined flow around a square cylinder were selected as unsteady flow examples to illustrate and analyse the properties of the parallel‐in‐time method through numerical experiments. The influence of several parameters on the computing time required to perform a parallel‐in‐time calculation on a PC cluster was verified. Among them we have analysed the influence of the number of processors, the number of iterations for convergence, the resolution of the spatial domain and the influence of the time‐step sizes ratio between the coarse and fine grids. Significant computer time saving was achieved when compared with the single processor computing time, particularly when the spatial dimension of the problem is low and the temporal scale is large. Copyright © 2004 John Wiley & Sons, Ltd.     

Riemann surface of dispersion diagram of a multilayer acoustical waveguide

A Riemann surface of a dispersion diagram of a multilayer planar waveguide is studied. Structure of this surface is linked with the transient wave phenomena in the waveguide. A method is proposed to study the dispersion diagram. The method is constituted in introducing of an artificial parameter controlling the links between the layers of the waveguide. The strength of the links varies from the full isolation to the realistic physical contact. The dependence of the solution on the artificial parameter is analytical. Using this approach it is proven that the branch points of the Riemann surface can be listed as the crossing points of the branches of non-interacting layers. A numerical procedure of finding the branch points positions is described. An explanation of the types of waves corresponding to different parts of the Riemann surface of the dispersion diagram is given.     

应用并行PEST算法优化地下水模型参数

基于列文伯格-马夸尔特（Levenberg-Marquardt）算法的PEST参数优化程序具有寻优速度快、健壮性好的优点,在地下水模型参数优化研究中有许多成功的应用实例。但是,对于大尺度、高精度和高复杂性的大规模地下水模拟,使用PEST进行参数优化需要大量的计算时间,优化效率较低。本文应用OpenMP并行编程方法对PEST算法进行了并行化,使之可以在共享存储并行计算机上进行参数优化的并行计算。并将此方法应用于甘肃北山区域地下水模型的参数优化中,并行实验表明,使用并行化的PEST可以将地下水模型参数优化效率提高3.7倍。
     

Efficient second‐order time integration for single‐species aerosol formation and evolution

The dynamics of a single‐species aerosol composed of droplets in air is described in terms of nucleation, evaporation, condensation, and coagulation processes. We present a comprehensive overview of the Euler–Euler formulation, which gives rise to a model in which fast nucleation that initiates aerosol droplets co‐exists with comparably slow condensation. The latter process is responsible for the subsequent growth of the droplets. To accurately represent the dynamical consequences of the fast nucleation process, while retaining numerical efficiency, a new second‐order time‐integration method for the nucleation, evaporation, and condensation processes is proposed and analyzed. The new time‐integration method takes the form of a ‘corrected Euler forward’ method. It includes rapid nucleation bursts and their possible cessation within a time step Δt. If the current nucleation burst persists for longer than the next time step, it is included fully, whereas cessation of the nucleation burst within the next Δt implies corrections to the effective rates in the algorithm. The identification of these two situations corresponds to the physical mechanism by which nucleation of a supersaturated vapor is halted because of the progressing condensation onto the already formed droplets. The resulting time‐integration method is shown to be second‐order accurate in time, whereas the computational costs per time step were found to be increased by less than 25% compared with the Euler forward method. The new method is also applied in combination with advective transport of the aerosol forming vapor to investigate a front of rapid nucleation. Adopting robust first‐order upwinding for the spatial discretization, we arrive at a flexible method that shows an overall first‐order convergence in Δt. For the full, spatially dependent system motivated by an aerosol of water droplets in air, the computational benefits of the new time‐integration method over the Euler forward scheme, are a factor of about 10 improvements in accuracy at a given Δt and a similar factor in computing time when keeping the same level of accuracy. Copyright © 2013 John Wiley & Sons, Ltd.     

A parallel adaptive numerical method with generalized curvilinear coordinate transformation for compressible Navier–Stokes equations

A fourth‐order finite‐volume method for solving the Navier–Stokes equations on a mapped grid with adaptive mesh refinement is proposed, implemented, and demonstrated for the prediction of unsteady compressible viscous flows. The method employs fourth‐order quadrature rules for evaluating face‐averaged fluxes. Our approach is freestream preserving, guaranteed by the way of computing the averages of the metric terms on the faces of cells. The standard Runge–Kutta marching method is used for time discretization. Solutions of a smooth flow are obtained in order to verify that the method is formally fourth‐order accurate when applying the nonlinear viscous operators on mapped grids. Solutions of a shock tube problem are obtained to demonstrate the effectiveness of adaptive mesh refinement in resolving discontinuities. A Mach reflection problem is solved to demonstrate the mapped algorithm on a non‐rectangular physical domain. The simulation is compared against experimental results. Future work will consider mapped multiblock grids for practical engineering geometries. Copyright © 2016 John Wiley & Sons, Ltd.     

基于有限元-时域分段自适应算法的围岩衬砌耦合渐进分析

以开挖时间、一衬时间、二衬时间划分时间区段,考虑围岩与衬砌为弹性/粘弹性介质,将围岩的“位移/应力释放不完全”、开挖等影响模拟为载荷的渐进释放,提出一个有限元-时域分段自适应算法,求解围岩衬砌耦合问题,并通过数值算例对算法进行了验证。所提方法给出了各时间区段的关联条件;可更准确描述各变量随时间的变化;同时将原时空耦合问题转化为一系列空间问题,并利用有限元方法递推求解;当步长变化时可通过自适应计算保证稳定的计算精度。此外,还建立了一个载荷渐进参数的反问题数值求解模型,并给出了相关算例。     

基于有限元-时域分段自适应算法的围岩衬砌耦合渐进分析

以开挖时间、一衬时间、二衬时间划分时间区段,考虑围岩与衬砌为弹性/粘弹性介质,将围岩的“位移/应力释放不完全”、开挖等影响模拟为载荷的渐进释放,提出一个有限元-时域分段自适应算法,求解围岩衬砌耦合问题,并通过数值算例对算法进行了验证。所提方法给出了各时间区段的关联条件;可更准确描述各变量随时间的变化;同时将原时空耦合问题转化为一系列空间问题,并利用有限元方法递推求解;当步长变化时可通过自适应计算保证稳定的计算精度。此外,还建立了一个载荷渐进参数的反问题数值求解模型,并给出了相关算例。     

Modeling of anisotropic polydispersed composites by progressive micromechanical approach

A progressive micromechanical method is presented in order to predict the elastic constants of polydispersed composites including multi-directional or randomly oriented reinforcement particles. Heterogeneities of various types are introduced into the matrices in a gradual manner. At each step, the Mori-Tanaka method is used to obtain the stiffness tensor of the intermediate medium used as a matrix of the following step. The proposed method is capable of introducing any kind of heterogeneities based on their dimensions, orientations, mechanical properties, and volume fractions to the matrix. Furthermore, suitable probability density functions can be defined for physical and structural parameters of the composite, including the level of the filler-matrix interfacial bonding, the aspect ratio, and the orientation of reinforcement particles. The efficiency of the iterative approach and the convergence of the solution are studied by computing the stiffness tensors of unidirectional and bidirectional particulate composites. The results of the present study are also compared with the literature data for a randomly oriented particulate composite.     

Use of two‐step split‐operator approach for 2D shallow water flow computation

The objective of this paper is to present a methodology of using a two‐step split‐operator approach for solving the shallow water flow equations in terms of an orthogonal curvilinear co‐ordinate system. This approach is in fact one kind of the so‐called fractional step method that has been popularly used for computations of dynamic flow. By following that the momentum equations are decomposed into two portions, the computation procedure involves two steps. The first step (dispersion step) is to compute the provisional velocity in the momentum equation without the pressure gradient. The second step (propagation step) is to correct the provisional velocity by considering a divergence‐free velocity field, including the effect of the pressure gradient. This newly proposed method, other than the conventional split‐operator methods, such as the projection method, considers the effects of pressure gradient and bed friction in the second step. The advantage of this treatment is that it increases flexibility, efficiency and applicability of numerical simulation for various hydraulic problems. Four cases, including back‐water flow, reverse flow, circular basin flow and unsteady flow, have been demonstrated to show the accuracy and practical application of the method. Copyright © 1999 John Wiley & Sons, Ltd.     

构建面向对象的并行有限元计算集群

介绍了一种面向对象的方法来组建计算机集群进行并行有限元分析。将构成集群系统的软、硬件组件设计成不同的类层次结构,形成一个完整的集群系统类库,并以系统分析、总体设计、结点通信以及系统效率测试等过程尝试以系统化的方法来分析整个构建过程。实际结果及并行有限元分析算例表明,这样的设计策略,可以非常清晰地从整体上把握系统的实现,大大降低系统构建的复杂性,使所生成的系统具有很好的可扩展性。