基于阶段识别的诺西肽发酵过程软测量建模

Phase identification based soft sensor modeling in Nosiheptide fermentation process

诺西肽是一种含硫多肽类抗生素,该抗生素是一种优良的非吸收型饲料添加剂,它能促进动物生长且在动物体内无残留。针对诺西肽发酵过程中关键生化参数难以在线测量的问题,提出了一种基于阶段识别的软测量建模方法。利用诺西肽发酵过程的非结构模型状态方程,根据隐函数存在定理确定出辅助变量,并利用K均值聚类算法进行阶段识别,根据识别结果对现场数据进行分类,然后采用多个神经网络分别构建出对应于各个阶段的局部软测量模型。实验结果验证了所提方法的有效性。

Nosiheptide, a novel type of sulfur-containing peptide antibiotics, is a perfect un-assimilated feed additive, which can promote animal growth without residual in animal body. A soft sensor modeling method based on phase identification is proposed to solve the difficulties of crucial biochemical parameter on-line measurement in Nosiheptide fermentation process. Firstly, using the state equations established for Nosiheptide fermentation process, the secondary variables are selected according to the implicit function existence theorem. Then, k-means clustering algo-rithm is used for phase identification, and the field data are classified according to the identification results. Lastly, a soft sensor model is developed, which consists of multiple local neural network models, one for each phase. Test results show the effectiveness of the presented approach.