多体动力学优化的并行SQP方法  

Optimization of Multibody Dynamics with Parallel SQP Algorithm

作　　者：刘小蒙[1] 魏继卿[1] 鲁玉祥[1] 

机构地区：[1]西北机电工程研究所,陕西咸阳712099

出　　处：《计算机仿真》2015年第5期290-292,301,共4页Computer Simulation

摘　　要：并行计算的发展大大提高计算机的计算效率,降低计算时间。针对多体动力学的优化问题,分析了求解灵敏度的三种方法的并行性,建立了有限差分法与直接微分法的并行算法。同时采用并行Armijo线性搜索,构成了完整的并行序列二次规划(SQP)算法。将上述算法应用到曲柄滑块的优化中,并与串行SQP算法进行了比较,证实了并行SQP算法可以大大降低计算时间。上述研究为多体动力学优化提供了一种并行求解思路。Development of Parallel computation consumedly raises the calculation efficiency and lower the calculation time. In the muhibody dynamics optimization issues, the parrel of three kinds of methods solving the sensitivity is analyzed, and the parallel algorithm of finite difference and direct differentiation is established. Parallel Armijo line search is adopted, and the major parallel SQP algorithm is constituted. The algorithm is used in the crank-slider mechanism, and is compared with traditional SQP algorithm. The result indicates that parrlel SQP algorithm can con- sumedly lower the calculation time.

关 键 词：并行 序列二次规划 直接微分 有限差分 

分 类 号：O313[理学—一般力学与力学基础]