An evolutionary deep learning soft sensor model based on random forest feature selection technique for penicillin...

Area of Science:

• Biotechnology
• Chemical Engineering
• Computational Science

Background:

• Accurate monitoring of biological parameters is crucial for optimizing penicillin production.
• Existing methods may lack the precision required for real-time fermentation control.
• Soft sensor models offer a promising approach for indirect measurement of key fermentation variables.

Purpose of the Study:

• To develop and validate a novel hybrid soft sensor model for penicillin fermentation.
• To improve the accuracy and reliability of biological parameter measurements during fermentation.
• To enhance the overall efficiency and yield of penicillin production through advanced process monitoring.

Main Methods:

• Feature selection using Random Forest (RF) to identify relevant auxiliary variables.
• An improved Harris Hawks Optimization (HHO) algorithm incorporating elite opposition-based learning (EOBL) and Golden Sine Algorithm (Gold-SA) for enhanced convergence and diversity.
• Construction of a Long Short-Term Memory (LSTM) network to build the soft sensor model.
• Validation using the Pensim simulation platform.

Main Results:

• The proposed RF-IHHO-LSTM soft sensor model demonstrated high measurement accuracy.
• The model effectively captured key biological parameters during simulated penicillin fermentation.
• The simulation results confirmed the model's suitability for practical engineering applications.

Conclusions:

• The hybrid RF-IHHO-LSTM soft sensor model provides an accurate and reliable method for monitoring penicillin fermentation.
• This advanced monitoring technique can significantly contribute to improving penicillin production efficiency.
• The developed model meets the stringent requirements of industrial penicillin fermentation processes.