Advanced Centre for Biochemical Engineering, Department of Chemical and Biochemical Engineering University College London, Torrington Place, London WC1E 7JE.
Biotechnol Bioeng. 1997 Jan 5;53(1):58-70. doi: 10.1002/(SICI)1097-0290(19970105)53:1<58::AID-BIT9>3.0.CO;2-Y.
Downstream processing operations are often carried out blind in the process timescale since product monitoring on-line is not common. Knowledge of the location and concentration of the product and key contaminants is complementary to other process information for process development and, if available on-line in conjunction with a suitable model, control. This article sets out to demonstrate a model describing a two-cut fractional protein precipitation process and how this may be used for control of the process to maximize yield in the face of variable process stream conditions. Estimation of the model parameters is achieved by means of data-fitting by least squares and in comparison prediction by a Kalman filter algorithm. A description and error analysis of equipment for at-line monitoring of the soluble product in a pilot plant environment is presented which includes a micro-centrifuge necessary to clarify small volumes of sample prior to analysis. Finally, an account of the successful implementation of this equipment and the Kalman filter algorithm for control at bench scale is given where conditions in the process stream are deliberately disturbed to test the control operation. (c) 1997 John Wiley & Sons, Inc.
下游加工操作通常在过程时间尺度上进行盲目操作，因为产品在线监测并不常见。了解产品和关键污染物的位置和浓度对于过程开发以及如果与合适的模型一起在线提供的其他过程信息是互补的。本文旨在演示一个描述两阶段分数蛋白沉淀过程的模型，以及如何在面对可变过程流条件时使用该模型来控制过程以最大化产量。通过最小二乘法的数据拟合和卡尔曼滤波算法的预测来实现模型参数的估计。介绍了一种在中试环境下在线监测可溶性产品的设备的描述和误差分析，其中包括在分析之前澄清小体积样品所需的微离心机。最后，给出了在台架规模上成功实施该设备和卡尔曼滤波算法进行控制的情况，其中故意干扰过程流中的条件以测试控制操作。（c）1997 年 John Wiley & Sons, Inc.
