| Abstrakt: |
Hyaluronic acid (HA), a glycosaminoglycan polymer, is widely used in the biomedical and cosmetic industries. Due to highly viscous nature of the HA fermentation broth, it is difficult to capture the dynamics of its bioprocess with physical sensors in real-time. The goal of this study was to track non-invasively the state variables involved in HA production process and deducing the critical process parameters based on the recorded process inputs. The framework employed in this study is based on a hybrid model that predicts HA and biomass concentration using online bioreactor data (pH, DO%, %CO2 evolved, feed rate, and agitation rate) to ensure real-time tracking of HA bioprocess dynamics. A HA fermentation dataset with data from historical batches and freshly performed fed-batch runs for various specific growth rate set-points (μsp) was created. The dataset was used to train the hybrid model, which was then used to predict biomass and HA concentration for test runs, with a mean squared error of prediction ranging from 0.018 to 0.049 (g/L)2. Furthermore, recurrent neural networks were evaluated in forecasting the specific growth rate (μ) and HA productivity rate (qHA) to observe the desired process trajectory. The current study addressed the scope of application of hybrid model based soft-sensor to predict the trend of process parameters of HA fermentation. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink