Identifyability measures to select the parameters to be estimated in a solid-state fermentation distributed parameter...

Area of Science:

• Chemical Engineering
• Biotechnology
• Process Systems Engineering

Background:

• Process modeling offers advantages in process control, cost reduction, and product quality improvement.
• Accurate models require precise parameter estimation.
• Solid-state fermentation (SSF) processes benefit from robust modeling for optimization.

Purpose of the Study:

• To develop and validate a distributed parameter model for solid-state fermentation.
• To perform parameter estimation and identifyability analysis (PIA) for model optimization.
• To assess the impact of substrate inhibition and parameter identifiability on model accuracy.

Main Methods:

• Development of a seven-differential-equation distributed parameter model for SSF.
• Application of parameter estimation techniques coupled with parameter identifyability analysis (PIA).
• Statistical validation of the model under different assumptions, including substrate inhibition.

Main Results:

• The model incorporating substrate inhibition demonstrated superior representation of the SSF process.
• Parameter identifyability analysis revealed eight non-identifiable parameters among the initial seventeen.
• Excluding non-identifiable parameters significantly improved model accuracy and estimation efficiency.

Conclusions:

• Parameter identifyability analysis is a valuable tool for refining process models by reducing parameter count.
• PIA enhances model accuracy and simplifies the estimation procedure, leading to more reliable process insights.
• The refined SSF model provides a more accurate representation for improved process control and optimization.