多体系统传递矩阵法分布式并行计算研究  

Study on distributed parallel computing of transfer matrix method for multibody systems

作　　者：杨海根[1] 芮筱亭[1] 刘怡昕[1] 张建书[1] 何斌[2] 

机构地区：[1]南京理工大学发射动力学研究所,江苏南京210094 [2]南京工业大学力学部,江苏南京211816

出　　处：《振动工程学报》2014年第1期9-15,共7页Journal of Vibration Engineering

基　　金：高校博士点基金资助项目(20113219110025);国家自然科学基金资助项目(10902051;11102089)

摘　　要：多体系统传递矩阵法作为一种多体系统动力学新方法,在进行多体系统动力学分析时无需系统总体动力学方程、涉及矩阵阶次低计算速度快的优点,已被广泛用于各种多管火箭、自行火炮、坦克等复杂大型机械系统动力学分析与设计。为了提高多体系统传递矩阵法计算速度,研究了分布式并行计算的原理和方法,将MPI(Message Parallel Interface)分布式并行计算环境引入到多体系统传递矩阵法计算中。针对变截面梁的多体系统传递矩阵法动力学计算算例,设计了并行算法,分析了单机多核和分布式网络环境对计算时间的影响,结果显示通过并行计算能够大幅度提高计算效率。As a new method for multibody system dynamics,the transfer matrix method has the following characteristics：dispensing with the global dynamics equation of system,low-order matrix involved,possessing high efficiency.Therefore,it has been widely used in the dynamic design of many complex mechanical systems,such as multiple-launch rocket system,self-propelled artillery system and tank system.In order to 1mprove the computing efficiency of this method,the MPI （Message Parallel Interface） distributed parallel computing environment is introduced into the transfer matrix method.A new parallel algorithm using transfer matrix method is designed to calculate the dynamics of non-uniform continuous beams,with the effect of the multi-core and distributed environment on the computing efficiency and acceleration ratio investigated.The results show that the parallel algorithm largely increases the computing efficiency of the transfer matrix method.

关 键 词：多体系统动力学 并行计算 分布式计算 多体系统传递矩阵法 MPI 

分 类 号：O313.7[理学—一般力学与力学基础]