基于支持向量机和模糊推理的毕赤酵母发酵过程故障诊断

Fault Diagnosis for Pichia pastoris Fermentation Based on a Hybrid Support Vector Machine and Fuzzy Reasoning Technique

摘要:

利用甲醇营养型毕赤酵母发酵生产外源蛋白过程中, 诱导期培养基中甲醇浓度波动严重影响发酵过程的稳定性。为此提出了基于智能型模式识别的发酵故障诊断系统。在某一时间窗口内, 利用支持向量机分类器对可在线测量的过程参数(发酵时间、搅拌转速、甲醇流加速率、O2消耗速率OUR、CO2释放速率CER)进行状态分类, 并结合模糊推理技术将发酵状态划分成甲醇浓度"适中"、"过量"、"匮乏"的3种模式, 构建智能型的故障诊断系统。在甲醇浓度不当的情况下, 该系统均能准确地判别出故障类型。通过补加甲醇或甘油, 可以对"错误"发酵批次进行补救。

Abstract:

In heterologous protein production by methylotrophic Pichia pastoris, the methanol concentration variation in culture medium severely affects fermentation stability. An intelligent pattern recognition based fault diagnosis system was thus proposed to solve the problem. A support vector machine classifier(SVM) was firstly used to categorize the on-line measurable parameters(fermentation time, agitation rate, methanol consumption rate, O2 uptake rate OUR and CO2 evolution rate CER) within a moving-window, and then the SVM was combined with fuzzy reasoning technique to construct an unique intelligent fault diagnosis system, which could classify the fermentation physiological states into 3 catagories of methanol "in shortage", "medium", and "in excess". In the cases of improper methanol concentration, the system could accurately identify the faults and their type. Then those failure-likelihood fermentations could be rescued by adding methanol or glycerol.