Stochastic identification of bioreactor process exhibiting input multiplicity

Area of Science:

• Biochemical Engineering
• Control Systems Engineering
• Non-linear Dynamics

Background:

• Bioreactor systems are inherently non-linear, posing challenges for traditional control strategies.
• Accurate mathematical models are crucial for developing effective model-based controllers.
• Continuous bioreactors exhibit input multiplicity, complicating modeling efforts.

Purpose of the Study:

• To address stochastic modeling challenges in bioreactor processes.
• To evaluate advanced non-linear modeling techniques for capturing complex system dynamics.
• To identify suitable methods for representing the characteristic bell-shaped steady-state curves.

Main Methods:

• Stochastic modeling based on input/output time-series data.
• Application of block-oriented Nonlinear AutoRegressive Moving Average with eXogenous inputs (NARMAX) structure.
• Utilizing Bootstrap structure detection for NARMAX model identification.
• Employing Wavelet-NARMAX models for enhanced system representation.

Main Results:

• The study investigates the efficacy of three distinct stochastic modeling techniques.
• These methods aim to accurately represent the non-linear characteristics, including input multiplicity and parabolic steady-state curves.
• The performance of block-oriented NARMAX, Bootstrap-NARMAX, and Wavelet-NARMAX are comparatively assessed.

Conclusions:

• Stochastic modeling approaches are vital for effectively modeling non-linear bioreactor systems.
• The evaluated techniques offer potential solutions for capturing complex dynamics and improving process control.
• Further research can build upon these methods for advanced bioreactor control strategies.