M.N. Cruz Bournazou

Online optimal experimental re-design in robotic parallel fed-batch cultivation facilities: Online Optimal Experimental Re-Design in Robotic

We present an integrated framework for the online optimal experimental re‐design applied to parallel nonlinear dynamic processes that aims to precisely estimate the parameter set of macro kinetic growth models with minimal experimental effort. This provides a systematic solution for rapid validation of a specific model to new strains, mutants, or products. In biosciences, this is especially important as model identification is a long and laborious process which is continuing to limit the use of mathematical modeling in this field. The strength of this approach is demonstrated by fitting a macro‐kinetic differential equation model for Escherichia coli fed‐batch processes after 6 h of cultivation. The system includes two fully‐automated liquid handling robots; one containing eight mini‐bioreactors and another used for automated at‐line analyses, which allows for the immediate use of the available data in the modeling environment. As a result, the experiment can be continually re‐designed while the cultivations are running using the information generated by periodical parameter estimations. The advantages of an online re‐computation of the optimal experiment are proven by a 50‐fold lower average coefficient of variation on the parameter estimates compared to the sequential method (4.83% instead of 235.86%). The success obtained in such a complex system is a further step towards a more efficient computer aided bioprocess development. Biotechnol. Bioeng. 2017;114: 610–619.

Online optimal experimental re‐design in robotic parallel fed‐batch cultivation facilities

We present an integrated framework for the online optimal experimental re‐design applied to parallel nonlinear dynamic processes that aims to precisely estimate the parameter set of macro kinetic growth models with minimal experimental effort. This provides a systematic solution for rapid validation of a specific model to new strains, mutants, or products. In biosciences, this is especially important as model identification is a long and laborious process which is continuing to limit the use of mathematical modeling in this field. The strength of this approach is demonstrated by fitting a macro‐kinetic differential equation model for Escherichia coli fed‐batch processes after 6 h of cultivation. The system includes two fully‐automated liquid handling robots; one containing eight mini‐bioreactors and another used for automated at‐line analyses, which allows for the immediate use of the available data in the modeling environment. As a result, the experiment can be continually re‐designed while the cultivations are running using the information generated by periodical parameter estimations. The advantages of an online re‐computation of the optimal experiment are proven by a 50‐fold lower average coefficient of variation on the parameter estimates compared to the sequential method (4.83% instead of 235.86%). The success obtained in such a complex system is a further step towards a more efficient computer aided bioprocess development. Biotechnol. Bioeng. 2017;114: 610–619.

Online optimal experimental re‐design in robotic parallel fed‐batch cultivation facilities for validation of macro‐kinetic growth models using E. coli as an example

We present an integrated framework for the online optimal experimental re‐design applied to parallel nonlinear dynamic processes that aims to precisely estimate the parameter set of macro kinetic growth models with minimal experimental effort. This provides a systematic solution for rapid validation of a specific model to new strains, mutants, or products. In biosciences, this is especially important as model identification is a long and laborious process which is continuing to limit the use of mathematical modeling in this field. The strength of this approach is demonstrated by fitting a macro‐kinetic differential equation model for Escherichia coli fed‐batch processes after 6 h of cultivation. The system includes two fully‐automated liquid handling robots; one containing eight mini‐bioreactors and another used for automated at‐line analyses, which allows for the immediate use of the available data in the modeling environment. As a result, the experiment can be continually re‐designed while the cultivations are running using the information generated by periodical parameter estimations. The advantages of an online re‐computation of the optimal experiment are proven by a 50‐fold lower average coefficient of variation on the parameter estimates compared to the sequential method (4.83% instead of 235.86%). The success obtained in such a complex system is a further step towards a more efficient computer aided bioprocess development. Biotechnol. Bioeng. 2017;114: 610–619.

Optimization of a Sequencing Batch Reactor process for waste water treatment using a two step nitrification model

Abstract In this work, Model Based Optimization (MBO) of Sequencing Batch Reactor (SBR) for Waste Water Treatment (WWT) with Activated Sludge Process (ASP) is presented. The importance of considering the nitrate bypass nitrification denitrification process as well as the nonlinear behavior of the process is presented. Moreover the optimization results, solving a nonlinear problem with an SQP algorithm and using a reduced 8 state model for the simulation of the process, are exposed. Accordingly, important improvements in operation strategies can be achieved in SBR process through model based optimization. In addition it be shown that the consideration of the nitratae bypass process is crucial when optimizing WWT plants.

A Novel Approach to Mechanism Recognition in Escherichia Coli Fed-Batch Fermentations

Abstract In this work, a novel systematic approach to achieve an efficient mechanistic modeling and simulation of fed-batch fermentations is presented. In order to show the efficiency of the developed simulation framework, data of Escherichia coli fed-batch fermentations are used. Fermentation processes are characterized by its dynamic behavior described by parameters such as growth rate, substrate concentration and cellular metabolic activity. Although there are models able to describe individual fed-batch fermentations, they become unreliable when fitted to new fermentations. To overcome this drawback, in this work different models are used at different optimal time points enabling not only a better description of the process, but also a better understanding of non measurable characteristics. By these means, three models compete in different intervals of the process. The candidate models are: an Overflow metabolism model (OF), a Citric Acid Cycle model (CAC) and a Survival or Maintenance model (M). Using an adequate model sequence, acetate formation, substrate consumption and cell growth are predicted with high accuracy. Moreover, the data needed to fit the models are reduced and a standardization of the model to be applied in different process states is enabled. Besides, with the development of a robust and effective model, the possibility of an online implementation for monitoring and control of the fermentation is exhibited. The results show that an efficient process monitoring based on the Dissolved Oxygen Tension and the Mechanistic Recognition is only limited by the convergence velocity of the algorithm.