Applying neural networks as software sensors for enzyme engineering.

Area of Science:

• Biotechnology
• Biochemical Engineering
• Process Control

Background:

• Enzyme production processes face challenges in online control due to inherent biological system variability.
• Lack of suitable online sensors for critical process variables hinders effective real-time monitoring and management.
• Accurate prediction of fermentation endpoints is crucial for economic optimization in enzyme manufacturing.

Purpose of the Study:

• To explore the application of artificial neural networks (ANNs) for computer-assisted control of enzyme production.
• To investigate the potential of ANNs as software sensors for key process variables in fermentation.
• To evaluate the feasibility of intelligent methods for predicting fermentation endpoints in enzyme production.

Main Methods:

• Development and training of feedforward-backpropagation neural network models.
• Utilizing trained neural networks as predictive software sensors for online process monitoring.
• Focus on factors influencing the performance of neural network models in enzyme production control.

Main Results:

• Demonstrated that well-trained feedforward-backpropagation neural networks can function effectively as software sensors.
• Indicated the potential for ANNs to provide real-time predictions for critical process variables.
• Highlighted that neural network performance is contingent upon various influencing factors.

Conclusions:

• Artificial neural network models present a viable approach for addressing online control challenges in enzyme production.
• ANNs can serve as effective software sensors, mitigating the need for specialized physical sensors.
• Intelligent control strategies based on ANNs hold significant economic promise for optimizing fermentation processes.