2种设计空间计算方法的比较研究——达标概率法和多指标叠加法

Comparison of two algorithms for development of design space-overlapping method and probability-based method

以党参提取工艺设计空间计算为例,比较了多指标叠加法和达标概率法等2种设计空间算法。采用模拟实验测定误差的方法计算达标概率,并且研究了模拟次数、计算步长和可接受达标概率阈值等参数对所得设计空间的影响。对于党参提取工艺的研究数据,模拟1万次,计算步长0.02可以获得满意结果。总体来看,多指标叠加法容易理解,多种商业软件具备此功能,但是未能说明在设计空间内操作时,工艺评价指标能达标的可靠程度。模拟实验误差的达标概率法计算相对复杂,但能通过可接受达标概率阈值体现设计空间保证工艺评价指标达标的可靠程度,且设计空间边缘不存在概率突变,因此更推荐达标概率法。

In this work,two algorithms（ overlapping method and the probability-based method） for design space calculation were compared by using the data collected from extraction process of Codonopsis Radix as an example. In the probability-based method,experimental error was simulated to calculate the probability of reaching the standard. The effects of several parameters on the calculated design space were studied,including simulation number,step length,and the acceptable probability threshold. For the extraction process of Codonopsis Radix,10 000 times of simulation and 0. 02 for the calculation step length can lead to a satisfactory design space. In general,the overlapping method is easy to understand,and can be realized by several kinds of commercial software without coding programs,but the reliability of the process evaluation indexes when operating in the design space is not indicated. Probabilitybased method is complex in calculation,but can provide the reliability to ensure that the process indexes can reach the standard within the acceptable probability threshold. In addition,there is no probability mutation in the edge of design space by probability-based method. Therefore,probability-based method is recommended for design space calculation.