Data Driven Modeling for Monitoring and Control of Industrial Fed-Batch Cultivations

Abstract

A systematic methodology for development of a set of discrete-time sequence models for batch control based on historical and online operating data is presented and investigated experimentally. The modeling is based on the two independent characteristic time dimensions of batch processing, being time within the batch and the batch number. The model set is parsimoniously parametrized as a set of local, interdependent models which are estimated from data for as few as half a dozen batches. On the basis of state space models transformed from the acquired input−output model set, the asymptotic convergence of iterative learning control is combined with the closed-loop performance of model predictive control to form an optimal controller aiming to ensure reliable and reproducible operation of the batch process. This learning model predictive controller may also be used for optimizing control through optimization of the bioreactor operations model. The modeling and preliminary control performance is demonstrated on an industrial fed-batch protein cultivation production process. The presented methods lend themselves directly for application as Process Analytical Technologies.

Challenges in Optimization and Control of Biobased Process Systems: An Industrial-Academic Perspective

“…First, datadriven modeling of industrial fermentation was achieved using partial least-squares 38 and subsequently time series modeling. 39,40 This was followed in refs 41 and 42 where the full power of first-principles modeling is exposed through a series of experimental validations. Furthermore, due to incomplete knowledge of underlying phenomena such as complex metabolic networks, hybrid modeling of biobased processes is a field with considerable potential.…”

Artificial neural network model and fuzzy logic control of dissolved oxygen in a bioreactor

“…Control of fed-batch bioprocess is challenging as accurate mathematical model is unavailable, also random variations occur in initial process state and model parameters from batch to batch. Lack of understanding on the microorganism's complex regulatory network and the interactions in bioreactor contributes to the absence of reliable mechanistic model [12]. An alternative to mechanistic model of penicillin fermentation is data-driven models using historical data of the process.…”

Monitoring Batch Processes Using Sparse Parallel Factor Decomposition

“…Most of the existing industrial approaches to ensure the consistency of batches are based on the precise sequencing and automation of all the stages in the batch operation . Batch processes, however, often suffer from the batch-to-batch disturbances such as variations in raw material properties, changes in the startup initialization, and disturbances encountered during an individual batch execution . Consequently, significant variations in the final product quality are often observed between batches .…”

“…1 Batch processes, however, often suffer from the batch-to-batch disturbances such as variations in raw material properties, changes in the startup initialization, and disturbances encountered during an individual batch execution. 2 Consequently, significant variations in the final product quality are often observed between batches. 2 Monitoring batch processes is therefore very important to ensure the safe operation and the production of consistent high quality products.…”

“…2 Consequently, significant variations in the final product quality are often observed between batches. 2 Monitoring batch processes is therefore very important to ensure the safe operation and the production of consistent high quality products. Batch process monitoring methods have focused on the use of fundamental mathematical models (e.g., theoretical models of batch processes), detailed knowledge-based models (e.g., expert systems or artificial intelligence methods), or data-driven models.…”