Study on Multi-Model Soft Sensor Modeling Method and Its Model Optimization for the Fermentation Process of Pichia...

Area of Science:

• Biotechnology
• Process Engineering
• Computational Biology

Background:

• Real-time measurement of key biomass variables in Pichia pastoris fermentation is challenging.
• Accurate monitoring is crucial for optimizing bioprocess yield and efficiency.

Purpose of the Study:

• To develop a novel multi-model soft sensor for accurate real-time biomass estimation in Pichia pastoris fermentation.
• To improve upon existing soft sensor modeling techniques for bioprocess monitoring.

Main Methods:

• A piecewise affine (PWA) multi-model approach was employed.
• Particle swarm optimization (PSO) with an improved compression factor (ICF) was used for model optimization.
• False nearest neighbor method determined PWA model order; least squares support vector machine defined local model scope.

Main Results:

• The proposed ICF-PSO-PWA multi-model soft sensor accurately captured nonlinear fermentation dynamics.
• Prediction accuracy was improved by 4.4884% compared to PSO-optimized weighted least squares vector regression.

Conclusions:

• The developed soft sensor provides a robust and accurate method for real-time biomass estimation in Pichia pastoris fermentation.
• This advancement offers significant potential for enhanced bioprocess control and optimization.