摘要
响应面法对提拉单晶生长参数进行数值模拟优化时,要通过单晶提拉的数值模拟计算获得响应面函数拟合所需要的试验数据,这些数据的获得占据整个响应面法优化求解的大部分时间。为提高优化效率,本文应用响应面方法的并行化算法对直拉单晶生长参数进行了优化。在Windows环境下搭建并行运算平台,采用C语言编制,以MPI消息传递方式的并行响应面优化并行程序。通过对串行和并行程序的优化所需时间进行对比,可以使并行后的优化效率提高80%以上。采用并行化的响应面法优化方法对提拉单晶生长向着更高的方向发展提供了更迅捷的途径。
To optimize Czochralski crystal growth parameters with response surface method, the necessary data for fitting response surface is to be obtained by numerical simulation, which would take most of the total optimization time. In order to enhance the efficiency of optimization, a parallel response surface method is proposed to optimize the single crystal growth parameter, with the parallel platform being built under the windows environment and the procedures being programmed with the MPI message transfer mode. Compared with the serial procedures, the parallel procedures enhances the efficiency by 80%. The parallel response surface method is proved to be an efficient method in the single crystal growth.
出处
《科技导报》
CAS
CSCD
2008年第22期50-53,共4页
Science & Technology Review
基金
北京市教委科研项目(KM200710005032)
关键词
熔体生长
响应面法
优化
并行计算
melt growth
response surface methodology
optimization
parallel computation
