Standardization of Process Norms in Baker's Yeast Fermentation through Statistical Models in Comparison with Neural...

Area of Science:

• Biotechnology
• Chemical Engineering
• Computational Biology

Background:

• Commercial yeast fermentation involves complex biochemical reactions, making consistent batch production challenging.
• Traditional methods struggle to model and control the dynamic, time-varying nature of these industrial processes.

Purpose of the Study:

• To develop a robust control system for time-varying yeast fermentation processes.
• To compare the efficacy of statistical modeling techniques against non-parametric artificial neural networks.

Main Methods:

• Utilized data from an industrial baker's yeast fed-batch fermentation process.
• Applied statistical methods, including projection pursuit regression.
• Employed non-parametric artificial neural network techniques, specifically backpropagation networks.

Main Results:

• Backpropagation neural networks exhibited the highest accuracy in predicting commercial yeast growth patterns.
• Projection pursuit regression also demonstrated significant prediction accuracy.
• Developed models aid in optimizing parameters to minimize yeast production variability.

Conclusions:

• Artificial neural networks, particularly backpropagation, offer a superior approach for modeling and controlling complex yeast fermentation.
• Statistical models provide valuable insights but are outperformed by neural networks for this application.
• The developed models can lead to more consistent and optimized commercial yeast production.