Distributed Algorithm for Computing the Vehicle Launch Dynamics under Interaction with the Medium

<p>The paper describes a distributed algorithm and a structure of the software package for its implementation in which a program for computing the vehicle launch dynamics under interaction with the medium flow is complemented with a program to determine the unsteady hydrodynamic loads by the v...

Full description

Bibliographic Details
Main Authors: G. A. Shcheglov, A. Yu. Shilov, E. V. Junak
Format: Article
Language:Russian
Published: MGTU im. N.È. Baumana 2015-01-01
Series:Nauka i Obrazovanie
Subjects:
Online Access:http://technomag.edu.ru/jour/article/view/358
Description
Summary:<p>The paper describes a distributed algorithm and a structure of the software package for its implementation in which a program for computing the vehicle launch dynamics under interaction with the medium flow is complemented with a program to determine the unsteady hydrodynamic loads by the vortex element method.</p><p>A distinctive feature of the developed system is that its local (running on a single computing core) LEAVING program to calculate the launch dynamics runs together with concurrent (running on multiple computing cores) MDVDD program to compute the unsteady vortex flow and hydrodynamic loads. The LEAVING program is the main one. It is launched app and then runs the MDVDD program in concurrent mode on the specified number of cores. Using MPI technology allows you to use a multiprocessor PC or a local network of multiple PCs to perform calculations. The equations of launcher spring-mass model dynamics and equations of vortex elements parameters evolution are integrated with the same time step. The interprogram communiaction in the step is provided asynchronously using the OS Windows Event mechanism (Events). Interfacing between LEAVING and MDVDD programs is built using the OS Windows FileMapping technology, which allows a specified data structure to be displayed and read to the fixed memory area.</p><p>The paper provides analysis of acceleration achieved with parallel processing on different numbers of cores, and defines a parallelization degree of various operations. It shows that the parallelization efficiency of the developed algorithm is slower than in case of calculation of the rigid body flow. The causes of reduced efficiency are discussed.</p><p>It is shown that the developed algorithm can be effectively used to solve problems on a small number of cores, e.g. on PC based on one or two quad-core processors.</p>
ISSN:1994-0408