CAJ | 학술논문

青霉素发酵过程具有明显的阶段特征，该特征从一些关键操作变量信息中能够得到反映。本文从反应过程的多个操作变量中，选取关键过程变量——冷水流加速率作为调度变量，并采用模糊C均值聚类算法对其进行分类，各聚类中心作为青霉素发酵过程的主要工况点；基于EM算法围绕不同工况点建立局部子模型，最后根据采样数据阶段特征的后验分布将各子模型融合。基于此方法采用Pensim仿真平台数据，能够辨识数据的阶段特征，并建立青霉素发酵过程的融合模型。仿真结果表明该模型具有较高的拟合精度，能对该发酵过程的主导变量进行比较精确的预测。
청매소발효과정구유명현적계단특정，해특정종일사관건조작변량신식중능구득도반영。본문종반응과정적다개조작변량중，선취관건과정변량——랭수류가속솔작위조도변량，병채용모호C균치취류산법대기진행분류，각취류중심작위청매소발효과정적주요공황점；기우EM산법위요불동공황점건립국부자모형，최후근거채양수거계단특정적후험분포장각자모형융합。기우차방법채용Pensim방진평태수거，능구변식수거적계단특정，병건립청매소발효과정적융합모형。방진결과표명해모형구유교고적의합정도，능대해발효과정적주도변량진행비교정학적예측。
Penicillin fermentation process has the feature of the distinct phases, which can be seen from some key operating variables. In this paper, the key process variable, namely, the cold water flow rate was taken as the scheduling variable, which was then classified by the fuzzy C-means clustering algorithm. The cluster centers were considered as the main operating points of the penicillin fermentation process. Local models were constructed around each operating point based on EM algorithm. Thereafter, sub-models were combined togethter according to the posterior distribution of the scheduling variable. The feasiblity and effectiveness of the proposed method was illustrated through the Pensim simulation platform.