3种关键工艺参数辨识方法的比较研究

Comparison on three methods for identification of critical process parameters

目的研究多元线性回归法、逐步回归法和标准偏回归系数法3种关键工艺参数辨识方法的特点。方法以黄芩水提液酸沉碱溶工艺关键参数辨识为例比较3种方法。结果 3种方法所得结果基本一致,即关键参数为药液质量分数、酸沉pH值、碱溶pH值和静置温度。结论 3种方法的阈值选择都有主观性;多元线性回归法思路最简单,但辨识关键工艺参数不如逐步回归法灵敏;标准偏回归系数法的优势在于能综合参数对多个指标的影响,并且易于加权计算。

Objective To investigate the characteristics of multiple linear regression method, stepwise regression method, and standardized partial regression coefficient method for the identification of critical process parameters. Methods The acid precipitation and alkaline dissolution process of Scutellariae Radix（SR） water extract was investigated as an example. Results Similar results were obtained using three different methods. The concentration of SR water extract, p H value of acid precipitation, p H value of alkali dissolution, and refrigeration temperature were considered as critical process parameters. Conclusion The selection of threshold values of the three methods all are subjective. Multiple linear regression method is simple, but is not as sensitive as stepwise regression method. The effects of a parameter on several process indices can be considered simultaneously and easily weighted using standardized partial regression coefficient method.