滑动轴承混合润滑多线程并行计算数值方法
发布时间：2018-07-27 17:18
【摘要】：针对混合润滑数值分析将动压效应、弹性变形和界面接触特性耦合而非常耗时的问题,基于共享内存并行系统的多线程程序设计语言OpenMP,提出一种多线程混合润滑并行计算数值方法——红黑线交叉并行计算法.该并行计算模型是将雷诺方程求解域分成两个相互独立的子求解域,依次对两个子求解域进行并行数值求解,可以有效克服CPU线程间数据争用问题,加快求解速度.着重研究了并行计算核数、网格数量和工作站配置对并行计算性能的影响,分析结果表明:并行计算模型能够有效提高滑动轴承混合润滑计算速度,并行计算速度的提升幅度与并行计算核数成非线性关系,随着CPU核数的增加计算速度的增加幅度逐渐减小;此外,与内存和缓存相比,CPU的主频对并行计算速度有非常大的影响.
[Abstract]:In view of the problem that the dynamic pressure effect, elastic deformation and interface contact characteristics are coupled by the numerical analysis of mixed lubrication, Based on OpenMP-based multithread programming language of shared memory parallel system, a numerical method of hybrid multithreading lubrication parallel computation, red and black line cross parallel computing method, is proposed. The parallel computing model divides the Reynolds equation into two independent subdomains, and performs the parallel numerical solution of the two subdomains in turn, which can effectively overcome the problem of data contention between CPU threads and speed up the solution. The effects of parallel computing kernel number, mesh number and workstation configuration on parallel computing performance are emphatically studied. The results show that the parallel computing model can effectively improve the speed of hybrid lubrication calculation of sliding bearings. The increase of parallel computing speed has a nonlinear relationship with the number of parallel computing kernels, and decreases gradually with the increase of the number of CPU kernels. In addition, the main frequency of CPU has a great influence on parallel computing speed compared with memory and cache.
【作者单位】： 重庆大学机械传动国家重点实验室;重庆大学汽车工程学院;四川大学空天科学与工程学院;
【基金】：国家自然科学基金资助项目(51375506,51475051)
【分类号】：TH133.31;TH117.2
，
本文编号：2148573
[Abstract]:In view of the problem that the dynamic pressure effect, elastic deformation and interface contact characteristics are coupled by the numerical analysis of mixed lubrication, Based on OpenMP-based multithread programming language of shared memory parallel system, a numerical method of hybrid multithreading lubrication parallel computation, red and black line cross parallel computing method, is proposed. The parallel computing model divides the Reynolds equation into two independent subdomains, and performs the parallel numerical solution of the two subdomains in turn, which can effectively overcome the problem of data contention between CPU threads and speed up the solution. The effects of parallel computing kernel number, mesh number and workstation configuration on parallel computing performance are emphatically studied. The results show that the parallel computing model can effectively improve the speed of hybrid lubrication calculation of sliding bearings. The increase of parallel computing speed has a nonlinear relationship with the number of parallel computing kernels, and decreases gradually with the increase of the number of CPU kernels. In addition, the main frequency of CPU has a great influence on parallel computing speed compared with memory and cache.
【作者单位】： 重庆大学机械传动国家重点实验室;重庆大学汽车工程学院;四川大学空天科学与工程学院;
【基金】：国家自然科学基金资助项目(51375506,51475051)
【分类号】：TH133.31;TH117.2
，
本文编号：2148573
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2148573.html
