Estimation of fungal biomass using multiphase artificial neural network based dynamic soft sensor

Area of Science:

• Biotechnology
• Bioprocess Engineering
• Artificial Intelligence in Biomanufacturing

Background:

• Cellulase production is crucial for biorefineries, but low-cost methods face challenges.
• Accurate biomass estimation is difficult in fed-batch fermentation with insoluble substrates.
• Controlling bioprocesses is hindered by the lack of direct biomass measurement.

Purpose of the Study:

• To develop a dynamic soft sensor for predicting fungal biomass concentration.
• To address the challenge of biomass estimation in the presence of insoluble substrates.
• To enhance the control and productivity of cellulase fermentation.

Main Methods:

• A Multiphase Artificial Neural Network (MANN) based dynamic soft sensor was developed.
• The MANN utilizes three Nonlinear Auto Regressive with eXogenous input (NARX) models for different fermentation phases (lag, log, stationary).
• Biomass concentration was predicted using online measurements like pH, substrate concentration, and agitation speed.

Main Results:

• The proposed MANN soft sensor demonstrated good performance in predicting biomass concentration.
• The model effectively captured the dynamic behavior of the bioreactor.
• Recursive biomass prediction showed acceptable deviation from real-time sensor data.

Conclusions:

• The developed MANN soft sensor offers a novel methodology for accurate fungal biomass estimation.
• This approach can lead to increased productivity of cellulase in industrial biorefineries.
• The soft sensor provides a viable solution for bioprocess control in complex fermentation media.