基于微粒群算法和网格模型的参数优化方法

Particle swarm optimization algorithm and grid model based method for parameter optimization

针对在半导体制造工艺参数优化过程中缺乏直观参考的问题,在微粒群优化算法(PSO)和等值线理论分析的基础上,将PSO与等值线矩形网格模型相结合,提出一种全新的工艺参数窗口选择方法,在二维标准多峰函数上验证了所提出方法的有效性,同时对所提出的方法进行了实际生产验证,对于双输入参数问题,该方法可以直接输出所有满足工艺要求的二维区域,从而为参数优化和范围选取提供直观参考,仿真测试结果和生产验证数据均表明了所提出的算法是一种有效的参数优化方法。

To effectively provide direct and clear reference for process parameter optimization in semiconductor manufacturing field, a new method to define input parameter window is proposed based on the advantages of both particle swarm optimization（PSO） algorithm and rectangle grid model. Simulation results on standard two-dimensional multi-modal functions show the effectiveness and validity of the method. For the scenario of two input parameters, the method can provide all the domains and a series of contour meeting process requirements. Both simulation results and production validation data show that the method is an effective way for process parameter optimization.