A study on optimal task decomposition of networked parallel computing using PVM

A numerical study is performed to investigate the effect of task decomposition on networked parallel processes using Parallel Virtual Machine (PVM), In our study, a PVM program distributed over a network of workstations is used in solving a finite difference version of a one dimensional heat equation, where natural choice of PVM programming structure would be the master-slave paradigm with the aim of finding an optimal configuration resulting in least computing time including communication overhead among machines. Given a set of PVM tasks comprised of one master and five slave programs, it is found that there exists a pseudo optimal number of machines, which does not necessarily coincide with the number of tasks, that yields the best performance when the network is under a light usage. Increasing the number of machines beyond this optimal one does not improve computing performance since increase in communication overhead among the excess number of machines offsets the decrease in CPU time obtained by distributing the PVM tasks among these machines. However, when the network traffic is heavy, the results exhibit a more random characteristic that is explained by the random nature of data transfer time.     

Branching of two-phase flow from a vertical header to horizontal parallel channels

The objective of the present experimental work is to investigate the two-phase flow distribution from a vertical main to parallel horizontal branches. Both the main and the branches have rectangular cross-sections simulating the header and the channels of the compact heat exchangers for air-conditioning systems. The cross section of the main is 8 mm × 8 mm while that of the parallel branches is 8 mm × 1 mm. Here, the second (downstream) junction was taken as the reference. The effect of the distance between the branches was mainly examined by changing it from 9 mm to 49 mm for the given flow conditions at the inlet of the downstream junction. Air and water were used as the test fluids. The superficial velocity ranges of air and water at the test section inlet were 13.2–21.4 m/s and 0.08–0.28 m/s, respectively. When the branch spacing becomes smaller, the fraction of liquid separation through the downstream branch decreases. The trend remains the same over the entire range of the present experiment, i.e., for different values of quality and the mass flow rate at the inlet of the downstream junction. Based on the correlation for single T-junctions, a modified correlation was proposed to take into account the effect of the branch distance in predicting the fraction of liquid separation. The correlation represents the experimental results within the accuracy of ±15 %. This paper was recommended for publication in revised form by Associate Editor Kyung-Soo Yang Jun Kyoung Lee received his B.S. degree in Mechanical Engineering from Busan National University in 1999. He then received his M.S. and Ph.D. degrees from KAIST in 2001 and 2005, respectively. Dr. Lee is currently a Professor at the School of Mechanical Engineering and Automation at Kyungnam University in Masan, Korea. Dr. Lee’s research interests are in the area of two-phase flow and heat transfer, micro-fluidics, cryogenic devices for superconductivity, thermal management systems for automobiles.     

一种基于经验模式分解的气液两相流流型识别方法

提出了一种基于经验模式分解的气液两相流流型识别方法.该方法首先对压差波动信号进行经验模式分解,将其分解为多个平稳的固有模式函数之和,再选取若干个包含主要流型信息的IMF分量,并从中提取时域特征指标-峭度系数作为LVQ神经网络的输入参数,从而实现流型的智能识别.对水平管内空气-水两相流流型的识别结果表明:以EMD为预处理器提取峭度系数的LVQ网络识别方法具有更高的识别率,可以准确、有效地识别流型.     

Adaptive parallel decomposition for multidisciplinary design

The conceptual design of a rotorcraft system involves many different analysis disciplines. The decomposition of such a system into several subsystems can make analysis and design more efficient in terms of the total computation time. Adaptive parallel decomposition makes the structure of the overall design problem suitable to apply the multidisciplinary design optimization methodologies and it can exploit parallel computing. This study proposes a decomposition method which adaptively determines the number and sequence of analyses in each sub-problem corresponding to the available number of processors in parallel. A rotorcraft design problem is solved and as a result, the adaptive parallel decomposition method shows better performance than other previous methods for the selected design problem.     

Numerical investigation on the transient characteristics of sediment-laden two-phase flow in a centrifugal pump

The objective of this work was to determine pressure fluctuation and transient flow characteristics, which aims to provide references to improve noise and vibration performance for the pump design and optimization when delivering sediment-laden flow. The three-dimensional (3D) transient simulations were simulated by SST k-ω turbulence model combined with Homogeneous equilibrium model (HEM). The experimental and numerical data was compared to validate the numerical accuracy. The simulation results predicted that the concentration shows strong effects on the external performance, velocity, pressure, turbulent kinetic energy distribution and peak amplitude of pulsation frequency, which all perform increasing trend with the rise of concentration. Meanwhile, the effect of the diameter size of particles on the flow field was relatively minor, which can also evidently influence the internal flow, but the effect is not simply proportional to the diameter size. The effect of diameter size on silt flow needs to be taken into account associated with the concentration distribution. The dominant frequency of solid-liquid approximately equals 0.8 times that of pure water, and the transient characteristics of sediment-laden flow perform low frequency with high amplitude features.     

反应堆周期监测装置的研制

本文比较详细介绍了反应堆周期装置的原理以及一种新型的电路的设计方法。经过在清华大学核能技术设计研究院屏蔽试验反应上两年的考验。证明该装置设计合理,性能稳定。     

并联机构雅可比矩阵的新式解法

介绍了并联机器人雅可比矩阵的两种新式解法：矢量构造法、基于符号运算的微分构造法。由于采用构造的方式计算雅克比矩阵，避免了以往对反解方程直接求导的计算方式，从而使雅克比矩阵的计算大为简化，另外利用计算机强大的计算功能，采用符号运算的方式也很方便。文中以6-PSS结构并联机器人为立，计算结果表明：这两种方法计算简单、准确、在工程应用中具有很强的实用性。     

基于经验模态分解和BP神经网络的油气两相流流型辨识

基于经验模态分解（empidcal mode decomposition，EMD）BP神经网络，提出了油气两相流流型辨识的新方法。应用EMD将差压信号分解成不同频率尺度上的单组分之和，并提取组分的归一化能量作为流型辨识特征量。BP神经网络以这些能量特征量为输入对油气两相流不同流型（包括泡状流、塞状流、层状流、弹状流和环状流）进行分类。实验结果表明，本文提出的流型辨识方法是有效的，其中泡状流、塞状流、层状流、弹状流和环状流的辨识精度分别为100％、89．4％,93．3％、96.3％和96.9％。     

A new decomposition method for parallel processing multi-level optimization

In practical designs, most of the multidisciplinary problems have a large-size and complicate design system. Since multidisciplinary problems have hundreds of analyses and thousands of variables, the grouping of analyses and the order of the analyses in the group affect the speed of the total design cycle. Therefore, it is very important to reorder and regroup the original design processes in order to minimize the total computational cost by decomposing large multidisciplinary problems into several multidisciplinary analysis subsystems (MDASS) and by processing them in parallel. In this study, a new decomposition method is proposed for parallel processing of multidisciplinary design optimization, such as collaborative optimization (CO) and individual discipline feasible (IDF) method. Numerical results for two example problems are presented to show the feasibility of the proposed method.     

A parallel computing application of the genetic algorithm for lubrication optimization

This study investigated the performance of parallel optimization by means of a genetic algorithm (GA) for lubrication analysis. An air-bearing design was used as the illustrated example and the parallel computation was conducted in a single system image (SSI) cluster, a system of loosely network-connected desktop computers. The main advantages of using GAs as optimization tools are for multi-objective optimization, and high probability of achieving global optimum in a complex problem. To prevent a premature convergence in the early stage of evolution for multi-objective optimization, the Pareto optimality was used as an effective criterion in offspring selections. Since the execution of the genetic algorithm (GA) in search of optimum is population-based, the computations can be performed in parallel. In the cases of uneven computational loads a simple dynamic load-balancing scheme is proposed for optimizing the parallel efficiency. It is demonstrated that the huge amount of computing demand of the GA for complex multi-objective optimization problems can be effectively dealt with by parallel computing in an SSI cluster.     

资源受限的机械振动 WSN 层次分解 CNN 边缘计算方法

用于机械振动监测的无线传感器网络节点的微控制器需要进行复杂的边缘计算,然而硬件资源受到限制。 卷积神经网 络作为一种性能优越的深度学习算法,若将其运行在 MCU 上可增强边缘 WSN 节点的计算能力。 本文提出了一种不修改 CNN 模型的层次分解方法,解决了难以在资源受限的 MCU 上运行不轻量化 CNN 的问题,实现了机械振动 WSN 节点的计算能力增 强。 首先通过设计文件结构用于分解并存储 CNN 模型参数,然后提出内存管理方法并推导随机存取存储器的消耗过程,最后 提出参数定位方法准确高效地读取模型参数。 实验表明仅使用 1. 76 KB RAM 与 2. 14 KB Flash,在 3. 15 ms 内便可实现高准确 率的边缘计算识别任务。     

两相流流动测试技术方法综述

两相流体的流动工况在动力、化工、核能、制冷、石油、冶金等工业中经常遇到。主要对流动的测试方法做了简单的总结，包括以两相流的流量、空泡率、流型、气泡各方面为测试对象的技术方法。     

Model analysis for differential pressure two-phase flow rate meter in intermittent flow

Multiphase flow rate metering is a challenging problem, specially for flow patterns other than wet-gas. This paper brings forward a new comparative analysis of three differential pressure calibration models suited for liquid dominated two-phase flows, in a total of seven model configurations. First, the models are compared theoretically and classified in terms of the type of input data required. Then, experimental data of over 300 horizontal air–water experiments, for $1$” and $2$” pipe diameters, supports quantitative analyses of the prediction accuracies and sensitivity of the superficial velocities of gas and liquid to measurement errors in the model input variables. Finally, a method for assessing the decoupled measurement errors for the void fraction and gas velocity is shown, as these variables are typically subject to higher uncertainties. It results that, though the void fraction is shown to be systematically under evaluated in more than 10%, the total mass flow rate is estimated through the Paz et al. (2010) model with an overall root mean squared deviation (RMSD) of 5.75% for the $2$” data. Also, the use of gas velocity measurements, even if subject to considerable errors, decreased the RMSD for the gas superficial velocity by more than half for the $1$” data.     

Coriolis flowmeter damping for two-phase flow due to decoupling

Coriolis flowmeters experience measurement errors due to both single- and two-phase flow. For two-phase flow, severe damping may occur, which leads to a (temporary) inability of the flowmeter to operate. The dominating part of the damping is caused by decoupling of the continuous and the dispersed phase. This paper presents the theory of damping due to decoupling in two-phase flow. Using a simple structural model, we provide examples of mixtures with water as the continuous phase. The dispersed phase is either air, or oil or sand.     

Miniature conductivity wire-mesh sensor for gas-liquid two-phase flow measurement

A miniature conductivity wire-mesh sensor for gas-liquid two-phase flow measurement in small channels is presented. The sensor design is similar to the conventional wire-mesh sensor for larger flow cross sections with wire electrodes stretched across the flow channel in two adjacent planes and with perpendicular wire orientation between planes. Conductivity measurement is performed by special electronics which consecutively applies bipolar voltage pulse excitation to the sender wires and measures electrical current flow in the wire crossings at the receiver wires. The new design is based on printed circuit board technology. A prototypical sensor made of 2×16 stainless steel wires each of 50 μm diameter was manufactured and applied to two-phase flow measurement inside the mixing chamber of an effervescent atomizer. Accuracy of the sensor was studied for static liquid distributions using microphotography and for dynamic two-phase flow by comparison of wire-mesh sensor data with radial gas fraction profiles obtained from X-ray microtomography measurements.     

气液两相流中上升气泡体积的计算方法

通过高速摄像机拍摄气液两相流中上升气泡的运动过程,分别记录不同直径的气孔所产生的单个气泡上升过程的连续图像,结合数字图像处理技术,提取了气泡的面积、当量直径、几何中心、速度、加速度等特征参数.并在此基础上,应用两种计算气泡体积的方法:几何方法和受力分析法,实现了气泡体积的初步计算.几何方法计算气泡体积是根据图像中气泡的形状构造一个与气泡体积相当的椭球体,进而计算椭球体的体积;而受力分析的方法则是通过分析气泡在垂直方向上的受力情况,根据牛顿第二定律推导出气泡体积的计算公式.并对两种方法计算出的气泡体积值进行分析对比,阐明了这两种方法适用情况,计算出单个气泡上升情况下的气相体积含气率.     

基于JAVA中的多线程技术实现并行计算

介绍了用Java中的多线程技术在多处理机上实现并行计算．以矩阵的乘法为例提出一个可行的算法并阐述其工作过程及特点。该算法实现了并行计算和Java中多线程技术之间的映射．为并行计算的实现提供了新思路。     

气液两相流型在线识别系统的开发和应用

在统计理论和小波理论的基础上,把神经网络应用到流型识别中,采用VB语言编程完成了软件的开发.按照信号测量、信号特征提取和流型识别3个实现流型客观自动识别的步骤,建立了一套气液两相流流型在线识别系统,在气液两相流实验系统上采集了180组仿真样本进行仿真实验.结果显示:流型正确识别率为91.7%,此系统应用于流型识别,不仅具有较高的识别率,而且很好地实现了识别结果的实时显示,运行速度快,达到了流型在线识别的目的.     

Liqnet: A real-time monitoring network for two-phase flow patterns

Real-time identification of gas-liquid two-phase flow can help fluid systems maintain safe operating conditions. A flow pattern identification method based on a convolutional neural network (CNN) algorithm (after this referred to as liqnet) is proposed in this paper to realize automatic detection and real-time identification of two-phase flow patterns. This paper mainly focuses on solving two problems of CNN algorithm flow pattern identification (1): the experimental samples for two-phase flow classification are few, and (2): the existing methods do not fully consider the real-time nature of two-phase flow identification. Therefore, this paper constructs a two-phase flow database containing 6242 images using data enhancement, proposes a lightweight network liqnet, and compares it with six mainstream CNN models. The results show that liqnet can achieve the highest accuracy (98.65%), has the least amount of parameters (1.3708 M), and can achieve the purpose of real-time prediction (32.11FPS).     

用于两相流流速测量的极性导向式自适应算法

提出利用导向式自适应算法处理两相流的流动噪声信号,以估计出的流动噪声传输参数计算两相流流速,对这种算法的收敛性和流速测量性能进行了论证.当流动噪声信号量化为1 bit时,则形成极性导向自适应算法,它便于用比较简单的硬件实现快速运算,适于制作成专用的流速检测集成芯片.理论分析和实验结果表明,这种方法测量结果的实时性和分辨率优于相关法.文中给出气水两相流垂直管段与极性相关法进行对比实验的结果.