Computer control of pH and DO in a laboratory fermenter using a neural network technique

Area of Science:

• Biochemical Engineering
• Process Control
• Artificial Intelligence in Biotechnology

Background:

• Maintaining stable pH and dissolved oxygen (DO) is crucial for biosystems control during fermentation.
• Automated control systems are essential for optimizing biochemical reactor performance.
• Saccharomyces cerevisiae fermentation requires precise monitoring of key process variables.

Purpose of the Study:

• To demonstrate the advantages of artificial neural networks (ANNs) for identifying and controlling biochemical reactors.
• To develop and implement a PC-supported, automated control system for laboratory-scale fermentation.
• To investigate the use of ANNs for maintaining critical parameters like pH and DO concentration.

Main Methods:

• Development of a PC-supported, automated, multi-task control system.
• Utilizing forward and inverse neural process models for system identification and control.
• Training neural models off-line using a modified back-propagation algorithm with conjugate gradients.
• Augmenting the inverse neural controller with a novel adaptive term for robust performance.

Main Results:

• Successful identification and control of pH and dissolved oxygen (DO) concentration in a Saccharomyces cerevisiae fermenter.
• Demonstration of robust performance due to the adaptive term in the inverse neural controller.
• Experimental validation of good regulatory and tracking performance for the proposed control system.

Conclusions:

• Artificial neural networks offer significant advantages for biochemical reactor control.
• The developed automated system provides effective and robust control of critical fermentation parameters.
• The study confirms the feasibility and efficacy of ANN-based control strategies in biosystems engineering.