Estimation and statistical analysis of model parameters using sequential Monte Carlo for phenol and p-cresol...

Area of Science:

• Chemical Engineering
• Separation Science
• Process Modeling

Background:

• Model-based design and optimization are vital for industrial chromatographic separations.
• Accurate quantification of model parameter uncertainty is essential for robust process design and control.

Purpose of the Study:

• To propose and evaluate a Bayesian approach using sequential Monte Carlo (SMC) for estimating parameter uncertainty in chromatographic models.
• To demonstrate the method's effectiveness using the linear driving force model for phenol and p-cresol separation.

Main Methods:

• Application of sequential Monte Carlo (SMC) methods within a Bayesian framework.
• Comparative analysis of different experimental injection tests (pulse injection and breakthrough experiments).
• Careful modeling of observation errors to ensure reliable parameter estimation.

Main Results:

• Confirmed the necessity of both pulse injection and breakthrough experiments for accurate and precise parameter estimation.
• Demonstrated that SMC effectively estimates parameter uncertainty for the linear driving force model.
• Highlighted the critical impact of accurately modeling observation errors on estimation quality.

Conclusions:

• The proposed Bayesian SMC approach provides a robust method for quantifying parameter uncertainty in chromatographic models.
• Optimal experimental design, including specific injection types and careful error modeling, is key to successful parameter estimation and reliable process design.