Fast inferential adaptive optimization of a continuous yeast culture based on carbon dioxide evolution rate

Abstract

A fast inferential, multivariable adaptive optimization algorithm based on a fast responding off-gas data, the carbon dioxide evolution rate (CER), has been developed and applied to a continuous baker's yeast culture to maximize the cellular productivity in simulation and experimental studies. In the simulation study the process was optimized based on CER measurements using readily available steady-state data on the ratio between the cellular productivity and the CER. It was shown that the algorithm is two to three times faster than the algorithm based on cell mass concentration measurements. In the experimental study the CER was maximized without any information on the relationship between the cellular productivity and the CER. It took about 40 h for the process to converge, while about 80 h was required when the optimization was based on cell mass measurements. The attained steady state was found to be different but fairly close to that obtained with cell measurements. Briefly discussed is a switching to the cell-mass-based algorithm at the final stage of the optimization to overcome a potential inaccuracy.