Random noise effects in pulse-mode digital multilayer neuralnetworks

A pulse-mode digital multilayer neural network (DMNN) based on stochastic computing techniques is implemented with simple logic gates as basic computing elements. The pulse-mode signal representation and the use of simple logic gates for neural operations lead to a massively parallel yet compact and flexible network architecture, well suited for VLSI implementation. Algebraic neural operations are replaced by stochastic processes using pseudorandom pulse sequences. The distributions of the results from the stochastic processes are approximated using the hypergeometric distribution. Synaptic weights and neuron states are represented as probabilities and estimated as average pulse occurrence rates in corresponding pulse sequences. A statistical model of the noise (error) is developed to estimate the relative accuracy associated with stochastic computing in terms of mean and variance. Computational differences are then explained by comparison to deterministic neural computations. DMNN feedforward architectures are modeled in VHDL using character recognition problems as testbeds. Computational accuracy is analyzed, and the results of the statistical model are compared with the actual simulation results. Experiments show that the calculations performed in the DMNN are more accurate than those anticipated when Bernoulli sequences are assumed, as is common in the literature. Furthermore, the statistical model successfully predicts the accuracy of the operations performed in the DMNN.     

Enhancing a tsunami evacuation simulation for a multi-scenario analysis using parallel computing

The numerical investigation of tsunami evacuation is becoming a major way to assess the potential evacuation risks and consider countermeasures, but it has been mostly limited to GIS-based static analysis or macroscopic agent-based modelling due to the costs of large-scale simulations. In this paper, we propose a simplified force-based evacuation simulation model and an easy-to-implement parallelization strategy for a large-scale microscopic tsunami evacuation simulation and demonstrate its applications in an actual urban environment. First, the simulation performance was verified and validated against experimental and observational results regarding basic pedestrian movement. The test results qualitatively and quantitatively showed good agreement with real pedestrian movements. The model was then applied to a case of tsunami evacuations in Kesennuma City, where the 2011 Tohoku tsunami caused devastating damage. The model was successfully scaled up to provide urban-scale characterization. In this application, the developed simulator was implemented by a hybrid MPI/OpenMP parallelized computing technique. By utilizing the proposed model with parallel computing, we achieved an urban-scale microscopic evacuation simulation five times faster than real-time and a stochastic simulation to evaluate the uncertainty in the evacuation simulation.     

Sampling-based RBDO using the stochastic sensitivity analysis and Dynamic Kriging method

This paper presents a sampling-based RBDO method using surrogate models. The Dynamic Kriging (D-Kriging) method is used for surrogate models, and a stochastic sensitivity analysis is introduced to compute the sensitivities of probabilistic constraints with respect to independent or correlated random variables. For the sampling-based RBDO, which requires Monte Carlo simulation (MCS) to evaluate the probabilistic constraints and stochastic sensitivities, this paper proposes new efficiency and accuracy strategies such as a hyper-spherical local window for surrogate model generation, sample reuse, local window enlargement, filtering of constraints, and an adaptive initial point for the pattern search. To further improve computational efficiency of the sampling-based RBDO method for large-scale engineering problems, parallel computing is proposed as well. Once the D-Kriging accurately approximates the responses, there is no further approximation in the estimation of the probabilistic constraints and stochastic sensitivities, and thus the sampling-based RBDO can yield very accurate optimum design. In addition, newly proposed efficiency strategies as well as parallel computing help find the optimum design very efficiently. Numerical examples verify that the proposed sampling-based RBDO can find the optimum design more accurately than some existing methods. Also, the proposed method can find the optimum design more efficiently than some existing methods for low dimensional problems, and as efficient as some existing methods for high dimensional problems when the parallel computing is utilized.     

基于Adams的列车碰撞三维动力学模型和仿真

为研究列车碰撞性能,用Adams创建由车体、车钩缓冲装置、端部吸能结构、防爬器、转向架和轮轨力等组成的单节车厢三维动力学模型,并创建6节车厢列车的三维动力学模型,模拟列车以15 m/s的速度与2节静止车厢碰撞的过程.通过分析各节车厢的速度、加速度和每个车钩缓冲装置的相对偏移量在碰撞过程中的变化情况,重现列车碰撞过程,进而分析影响列车垂向爬车和横向屈曲稳定性的因素.仿真结果表明,碰撞过程中每个车钩缓冲装置的相对偏移量和列车各节车厢的加速度最大值均沿列车运行向不断变小;列车前三节车厢的垂向爬车和横向屈曲最严重,转向架发生出轨现象.     

基于联合权重超图划分的SNN负载均衡方法

大规模脉冲神经网络并行模拟是探究大脑机能的重要手段。其难点在于合理地将负载映射到并行分布式平台上,提升模拟速度。为解决该问题,提出一种基于联合权重超图划分的SNN负载均衡方法,解决并行计算中进程间计算负载与通信负载的均衡问题,提高SNN模拟速度。并使用稀疏通信的方式替代集体通信,解决事件通信过程中的数据冗余问题,提升通信效率。实验结果表明,该方法使带有STDP突触20%规模的皮质层微电路模型的模拟时间,比标准循环分配算法缩短约64.5%,比普通超图分配算法缩短约57.4%,同时事件通信数据量减少了90%以上。     

基于分层抽样的模拟禁忌混合智能优化算法TSII

将分层抽样随机模拟与禁忌搜索结合,构造了TS II模拟禁忌混合智能优化算法。随机模拟采用缩减方差、加速收敛的分层抽样技术,保证抽样遍布于整个搜索空间,避免禁忌搜索路径往返重复,克服禁忌搜索对初始解的依赖,算法同时使用禁忌表与希望表,将分散搜索与集中搜索相结合,增强算法的并行处理能力,提高寻优的效率与精度。Benchmark问题评测结果显示出了该算法的有效性。     

Air pollution modelling using a Graphics Processing Unit with CUDA

The Graphics Processing Unit (GPU) is a powerful tool for parallel computing. In the past years the performance and capabilities of GPUs have increased, and the Compute Unified Device Architecture (CUDA) - a parallel computing architecture - has been developed by NVIDIA to utilize this performance in general purpose computations. Here we show for the first time a possible application of GPU for environmental studies serving as a basement for decision making strategies. A stochastic Lagrangian particle model has been developed on CUDA to estimate the transport and the transformation of the radionuclides from a single point source during an accidental release. Our results show that parallel implementation achieves typical acceleration values in the order of 80-120 times compared to CPU using a single-threaded implementation on a 2.33 GHz desktop computer. Only very small differences have been found between the results obtained from GPU and CPU simulations, which are comparable with the effect of stochastic transport phenomena in atmosphere. The relatively high speedup with no additional costs to maintain this parallel architecture could result in a wide usage of GPU for diversified environmental applications in the near future.     

The direct simulation Monte Carlo method applied to a Boltzmann-like vehicular traffic flow model

In this paper a direct simulation Monte Carlo (DSMC) method is applied to a spatial homogeneous mesoscopic vehicular traffic flow model, based on a Boltzmann-like master equation. In contrast to gas kinetics, where in a collision a velocity jump change occurs, the interaction now changes the acceleration value of the following car in a leading car pair. There are no conservation laws in a single interaction. Therefore the Bird simulation scheme seems not to be the right choice for the approximation of the interaction integral. It is shown that a Nanbu like scheme is natural for this process. To avoid the typical double loop computational effort of the Nanbu scheme, a sampling algorithm developed by Babovsky is applied. Several car interaction profiles are examined and their resulting stochastic equilibrium solutions are discussed. First, simple interaction profiles are used to compare the simulation results with analytic calculated velocity distributions showing excellent agreement. Second, a realistic distance threshold interaction profile is applied to the simulation and the results are shown to be in qualitative agreement with measured traffic flow data. The simulation procedure seems to be applicable to study the influence of different interaction profiles to the macroscopic vehicular traffic flow quantities in stochastic equilibrium.     

Parallel computing of 3D smoking simulation based on OpenCL heterogeneous platform

Open Computing Language (OpenCL) is an open royalty-free standard for general purpose parallel programming across Central Processing Units (CPUs), Graphic Processing Units (GPUs) and other processors. This paper introduces OpenCL to implement real-time smoking simulation in a virtual surgery training simulation system. Firstly, the Computational Fluid Dynamics (CFD) is adopted to construct the real-time smoking simulation model based on the Navier?CStokes (N-S) equations of an incompressible fluid under the condition of normal temperature and pressure. Then we propose a parallel computing technique based on OpenCL to accomplish the parallel computing of smoking simulation model on CPU and GPU, respectively. Finally, we render the smoke in real time by using a three-dimensional (3D) texture volume rendering method. Experimental results show that the parallel computing technique we have proposed achieve a satisfactory effect on image quality and rendering rate both on CPU and GPU.     

基于MPI的电网仿真实时并行计算平台

在电力系统动态电磁暂态仿真的并行计算中,存在超实时和硬实时的问题。为此,提出一种基于MPI的实时并行计算平台。引入硬实时操作系统RTLinux,采用实时内核和PSDD编程模式对仿真并行计算程序、MPICH并行环境和GM软件等进行重构,以获得硬实时特性。测试结果证明,该平台的平均时间性能提高约10%,时间的最大抖动幅度降低50%~80%,并减少了时间的抖动频率。     

成年树光照资源仿真的快速计算

光照是树木生长需要的重要资源,是树木生长仿真计算中必不可少的因素.但在森林演化的计算机模拟中,由于光照模拟的复杂性,使得光照模型的计算量十分巨大.本文采用了光照指数(Gap Light Index,GLI)为因子的光照模型,并针对该模型开展了快速计算研究.由于该模型计算中存在着大量限制计算效率的几何求交运算,本文根据光照指数与植物暴露面积所具有的共同特点,提出采用基于暴露面积的计算方法来近似拟合GLI的值,并在并行计算架构CUDA上实现了该算法.最后通过不同实验比较,验证了本文提出的方法针对较大的树木规模,在保证较小误差率的前提下,获得了比GPU并行求交方法快数十倍以上的加速比.     

一种孔隙介质中地下水流并行计算方法

针对孔隙介质中地下水流动问题提出了一种并行数值计算方法,并基于此设计了一套专用于求解大规模三维地下水流动方程的并行计算模块。计算模块基于区域分解的方法实现对模型区域的并行求解,采用了分布式内存和压缩矩阵技术解决大规模稀疏矩阵的存储及其计算,整合多种并行Krylov子空间方法和预条件子技术迭代求解大规模线性方程组。在Linux集群系统上进行了数值模拟实验,性能测试结果表明,程序具有良好的加速比和可扩展性。     

基于动态岛屿群体模型的并行遗传算法

在科学计算领域，并行计算越来越成熟，并行遗传算法开始受到关注，文章分析了遗传算法并行化的动机和实现模型，提出了一种新算法－基于动态岛屿群体模型的并行遗传算法，仿真结果验证了这种新算法的有效性和合理性。     

基于Docker的MPI和OpenMP混合编程

针对当前搭建集群并行系统复杂且耗时等问题,提出基于Docker搭建并行系统。介绍轻量级虚拟化技术Docker的核心概念和基本架构,并基于Docker技术在Linux平台上搭建集群并行开发环境。简要阐述并行计算的思想,叙述MPI和OpenMP并行计算的基本概念和特点,针对矩阵并行乘法的算法建立MPI和OpenMP的混合编程模型,并给出混合编程模型与MPI并行编程模型以及OpenMP并行编程模型的性能对比,分析出现差异的原因。基于该混合编程模型比较Docker与传统物理机两者搭建的并行系统的并行效率。     

YH-ACT：热工流体力学并行应用程序

商业CFD程序已广泛应用于反应堆的热工水力模拟,但不能完全满足反应堆的应用需求;开源CFD程序有部分应用,但与商业CFD程序相比,在物理模型全面性、计算精度、计算效率及易用性等方面仍存在差距。为更好地满足局部精细热工水力分析的需求,需要更全面的物理模型、较高的计算精度和较好的并行计算效率,因此有必要开发自主热工CFD程序。详细描述了热工流体力学并行应用程序YH-ACT的设计、实现方案以及测试结果。选取3个典型案例,通过与典型商业软件Fluent计算结果进行对比验证软件正确性, 程序并行计算规模达到400个结点共9 600个进程,稳态计算加速比为111.7,并行效率为27.9%,瞬态计算加速比为37.2,并行效率为9.3%。     

车组偏差系统的分散控制

研究汽车车组速度和距离偏差系统的控制问题。针对具有重叠结构的车组速度与距离偏差的线性随机系统模型 ,利用系统的包含原理 ,给出一种分散、次优的 L QG控制设计方法 ,并给出了利用该方法对车组偏差系统进行控制的仿真结果     

Multidisciplinary optimization of car bodies

Rising complexity of industrial development in the automotive industry is leading to a higher degree of interdisciplinarity, which is especially true in the virtual design area. New methods and solution procedures have to be evaluated and integrated in the overall process. For example, in car body design process, a new topic emerged recently: the multidisciplinary optimization of car bodies with respect to crash and NVH (noise, vibration, and harshness). Because rigorous evaluation of appropriate numerical algorithms is still missing, an intense study was realized at the research center of BMW. The results are summarized in this article. Four benchmarks have been studied: (a) a full vehicle model for NVH analysis, (b) a simplified multidisciplinary problem with a single crash case and linear statics and dynamics, (c) a lateral impact problem for multi-criteria optimization, and finally, (d) a small shape optimization problem was included to demonstrate the potential of transferring the results to the more complex problem of optimizations based on real changes in the shape of the structures. Because response surface methods have already been discussed in the literature and because of their failure in certain industrial cases, the focus was set on the evaluation of stochastic algorithms: simulated annealing, genetic and evolutionary algorithms were tested. Finally, a complete industrial multidisciplinary example from the current development process was studied for the validation of the results.     

汽车与行人碰撞的动力学响应仿真研究

行人保护是汽车安全研究领域的重要问题之一。该文旨在研究真实事故中汽车前碰撞行人的动态响应。基于这一点，使用了一个经过验证的行人数学模型，模拟真实碰撞事故中行人的动态响应。文中对行人模型的运动学响应以及真实事故记录进行了比较，并且计算了头、胸、骨盆、下肢等人体各部分与损伤相关的参数。基于模拟研究的结果，提出了减少行人损伤风险的城区限行时速和通过改进汽车前部结构减少行人碰撞损伤严重程度的可行措施。     

A Hybrid Analysis of an Optimization Approach for Cluster Applications

Cluster/distributed computing has become a popular, cost-effective alternative to high-performance parallel computers. Many parallel programming languages and related programming models have become widely accepted on clusters. However, the high communication overhead is a major shortcoming of running parallel applications on cluster/distributed computing environments. To reduce the communication overhead and thus the completion time of a parallel application, this paper introduces and evaluates an efficient Key Message (KM) approach to support parallel computing on cluster computing environments. In this paper, we briefly present the model and algorithm, and then analytical and simulation methods are adopted to evaluate the performance of the algorithm. It demonstrates that when network background load increases or the computation to communication ratio decreases, the analysis results show better improvement on communication of a parallel application over the system which does not use the KM approach.     

自适应的多种群并行遗传算法研究

典型的遗传算法本质上是一种并行的随机搜索策略,它不能对进化方向做出正确的感知和预测。文章定义了一个可以感知进化方向和衡量进化速度的指标,指导遗传算子做出自适应的调整。结合并行计算和多种群进化思想,提出了“带环多种群模型”,并构造了一种自适应的多种群并行遗传算法。实验结果表明,该算法可以有效地引导和控制进化方向,克服过早收敛现象,提高搜索效率。