Abstract: Through establishing a multi-variable relationship between process controlling parameters and detecting parameters,a process stability control model was proposed.Based on process physical laws or process statistical data,a multi-variable linear model between process controlling parameters and detecting parameters was established.Through parameter estimation,the prediction ellipsoid and joint confidence intervals of detecting parameters,as well as the expected value of multi-variable controlling parameters which met the detecting parameters stability requirements were acquired.In view of three commonly used objective functions in actual process,by considering model error and parameter fluctuations,the genetic algorithm was adopted to solve the fluctuation range of controlling parameters that meet the detecting parameters'stability requirements at a given confidence.The feasibility of the proposed method was illustrated by a typical manufacturing process of a product.

Key words: process stability, multi-variable linear model, prediction, control parameters, genetic algorithms

摘要： 为确保产品质量和可靠性水平,提出一种通过建立工艺过程控制参数与检测参数多变量线性关系来进行工艺稳定性控制的建模方法。以工艺物理规律或工艺统计数据为基础建立工艺过程控制参数与检测参数的多变量线性模型,通过参数估计得到检测参数的预测椭球和联合置信区间以及满足检测参数稳定性要求的多变量控制参数期望取值。针对实际工艺过程常用的三种目标函数,综合考虑模型误差与参数波动,采用遗传算法求解在给定置信度要求下满足检测参数稳定性要求的控制参数波动区间。以某产品典型制造工艺为例说明了所提方法的可行性。

关键词: 工艺稳定性, 多变量线性模型, 预测, 控制参数, 遗传算法