连续系统模型的分布并行仿真方法研究  

Research on Distributed Parallel Simulation Method of Continuous System Model

作　　者：范威威 姚新宇[1] 吴文波[1] 

机构地区：[1]国防科技大学信息系统与管理学院,湖南长沙410073

出　　处：《计算机仿真》2014年第2期344-348,449,共6页Computer Simulation

摘　　要：在算法实时性问题的研究中,工程系统多数是连续系统,以微分方程组的形式来表示的,通常采用数值积分算法来求解微分方程组。如果采用单一节点仿真,往往不能满足实时性要求,通常是将模型分割并分布至多节点进行分布并行仿真。针对Simulink仿真的机制和特点,基于RK4积分算法,将范德波尔模型正确分割成两个子系统SS1和SS2,以状态变量为输出变量,为使子系统间同步并行计算,分别在子系统的输出端配置Memory模块来优化分布仿真的效率。其它仿真软件如RT-LAB也基于Memory模块来分布仿真,但是没有分析Memory模块给系统动态带来的影响。在分析得出Memory模块的引入给系统带来了额外一拍时延的基础上,针对延误差提出拟合外推(Least squares estimator,LSQE)算法进行外推补偿,实验结果表明,相比较传统的插值外推(Lagrange,LAGRA)算法而言,拟合外推算法的补偿效果更好。Most of engineering systems are continuous systems which are represented as differential equations. Numerical integral algorithms are used to solve differential equations. It can not accomplish with requirement of real - time simulation in a single node. It is a common way to assign distributed model to multi - nodes for simulation. In this paper, we concentrate on the simulation mechanism and character of Simulink, based on RK4 integral algo- rithms, cut original model into two subsystems of SS1 and SS2 based on VDP （Vanderpol） equation, configure state variables to be output variables of SS1 and SS2 and then configure Memory block after outputs of SS1 and SS2 to en- hance efficiency of distributed simulation. Bring forward LSQE （Least squares estimator） algorithms to compensate for time -delay error, results of experiment show that LSQE （least squares estimator） algorithms are better than tra- ditional LAGRA （Lagrange） algorithms.

关 键 词：子系统 分布并行仿真 时延 

分 类 号：TP271.6[自动化与计算机技术—检测技术与自动化装置]