Soft - sensing modeling based on ABC - MLSSVM inversion for marine low - temperature alkaline protease MP...

Area of Science:

• Biotechnology
• Biochemical Engineering
• Process Control

Background:

• Marine low-temperature protease fermentation presents challenges due to nonlinear dynamics, multiple parameters, strong coupling, and difficult on-line measurement.
• Accurate monitoring of key biological parameters is crucial for optimizing fermentation efficiency and product yield.

Purpose of the Study:

• To develop a soft-sensing modeling method for real-time prediction of key biological parameters in marine protease fermentation.
• To address the limitations of direct on-line measurement in complex fermentation systems.

Main Methods:

• Established a dynamic "grey box" model of the fed-batch marine protease fermentation process.
• Developed an inverse model using Multiple Least Squares Support Vector Machine (MLSSVM) and optimized it with the Artificial Bee Colony (ABC) algorithm.
• Integrated the corrected inverse model with the fermentation process to create a composite pseudo-linear system for on-line prediction.

Main Results:

• The proposed soft-sensing modeling method successfully enabled real-time, on-line prediction of key biological parameters in the alkaline protease MP fermentation process.
• Demonstrated higher accuracy and better generalization ability compared to traditional Support Vector Machine (SVM) soft-sensing methods.

Conclusions:

• The developed method offers a new approach for soft-sensing modeling in fermentation processes, effectively solving the problem of real-time parameter prediction.
• The methodology can be extended to soft-sensing modeling of other general nonlinear systems, broadening its applicability.