Soft-sensor development for monitoring the lysine fermentation process.

Area of Science:

• Biotechnology and Biochemical Engineering
• Industrial Microbiology
• Process Analytical Technology

Background:

• Effective monitoring of fermentor processes is crucial for consistent, high-level production of target compounds.
• Real-time estimation of key parameters like cell growth, substrate, and product concentration is challenging with traditional methods.

Purpose of the Study:

• To develop a novel soft sensor for estimating lysine, sucrose, and bacterial cell concentrations in commercial fermentors.
• To leverage machine learning and available on-line process data for enhanced fermentation monitoring.
• To assess the applicability of the soft sensor across multiple fermentor tanks.

Main Methods:

• Development of soft sensor models using machine learning algorithms (linear and nonlinear).
• Utilization of on-line process data from commercial fermentors.
• Data enhancement techniques, including time interpolation, to improve model performance.
• Validation of the soft sensor across multiple fermentor systems.

Main Results:

• Accurate estimation of lysine concentration using both linear and nonlinear models.
• Successful estimation of sucrose and bacterial cell concentrations using nonlinear models.
• Improved prediction accuracy and reduced fluctuations through data enhancement via time interpolation.
• Effective estimation of fermentation behavior across multiple tanks using a single soft sensor model.

Conclusions:

• Machine learning-based soft sensors offer a robust solution for real-time monitoring of complex biotechnological processes.
• The developed soft sensor can accurately predict critical fermentation parameters, aiding in process stabilization and optimization.
• This technology represents a significant advancement towards the digital transformation of industrial fermentation.