Open AccessGeneric biomass estimation methods targeting physiologic process control in induced bacterial cultures
Author(s) -
Reichelt Wieland N.,
Thurrold Peter,
Brillmann Markus,
Kager Julian,
Fricke Jens,
Herwig Christoph
Publication year - 2016
Publication title -
engineering in life sciences
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.547
H-Index - 57
eISSN - 1618-2863
pISSN - 1618-0240
DOI - 10.1002/elsc.201500182
Subject(s) - bioprocess , soft sensor , biomass (ecology) , process (computing) , biochemical engineering , computer science , data mining , process engineering , process control , microbiology and biotechnology , biological system , engineering , biology , ecology , operating system , chemical engineering
Advanced bioprocess development strategies focus on the control of physiological entities, which rely on accurate real‐time determination of the biomass concentration. Various methods have been proposed in literature but up to this date a comprehensive and differentiated comparison of biomass estimation approaches for early stage bioprocess development is missing. In this study, we compared hard sensor, soft‐sensor, and data‐driven approaches for real‐time biomass estimation in respect to accuracy, transferability, and costs. The outlined methods were tested with two different microbial strains and recombinant products using Escherichia coli . To investigate the applicability of the outlined methods, method performance was assessed in correspondence to metabolic activity. Based on statistical descriptors the methods were compared and discussed. The results indicate no significant impact of strain or biomass estimation approach on the measurement quality. The average relative error of 11–13% can be greatly reduced by over 85% combining the outlined methods by the means of weighted average. This approach proved to be highly robust even during highly dynamic process conditions of oscillating specific substrate uptake rates. Concluding, the combination of low cost first principle soft‐sensor approaches in combination with a hybrid soft‐sensor yields the best information‐to‐effort ratio.