(300a) "Evolving Applications Of Chemical Engineering To Problems In Biology

The potential benefits from applying multivariable control to ensure reproducible operation of cultivations are substantial. Such technologies can be further exploited to enable Optimizing Control of cultivations to harvest additional benefits. The intentions behind the PAT initiative are clearly to improve the harvest by exploiting recent analytical and control theoretical advances. The aim of the present contribution is to demonstrate that recent advances can provide a substantial contribution to facilitating reproducible and optimizing operation of cultivations. These benefits are possible despite the fact that modelling of regulatory networks in microorganisms still is in its very early stage. Two principally different approaches to solving the supervision and control challenges are given. One approach is based upon operational data for soft sensor and for inductive statistical model development. A systematic methodology for acquiring discrete-time state space models for batch or continuously operating cultivations based on standard and/or designed operational data is introduced. This data driven model type may be used for monitoring and/or guiding the course of a cultivation to follow that of a previously satisfactory cultivation. The present contribution presents how the asymptotic convergence of Iterative Learning Control may be combined with the closed-loop performance of Model Predictive Control to form a robust and asymptotically stable optimal controller for ensuring reliable and reproducible operation of batch and periodic processes. Consequently this controller type may also be used for Optimizing control. The alternative approach is based upon a deductive process model to be used for model based monitoring and/or control design. In general first principles models for cultivations are still of very limited usability due to the largely unknown regulatory network in microorganisms however in specific cases special models may be developed and used for monitoring and control. Results from both approaches will illustrate their relative merits, including feasibility and ease for practical application on continuous and fed-batch cultivations for monitoring and control.