Uncertainty quantification via bayesian inference using sequential monte carlo methods for CO2 adsorption process

This work presents the uncertainty quantification, which includes parametric inference along with uncertainty propagation, for CO2 adsorption in a hollow fiber sorbent, a complex dynamic chemical process. Parametric inference via Bayesian approach is performed using Sequential Monte Carlo, a completely parallel algorithm, and the predictions are obtained by propagating the posterior distribution through the model. The presence of residual variability in the observed data and model inadequacy often present a significant challenge in performing the parametric inference. In this work, residual variability in the observed data is handled by three different approaches: (a) by performing inference with isolated data sets, (b) by increasing the uncertainty in model parameters, and finally, (c) by using a model discrepancy term to account for the uncertainty. The pros and cons of each of the three approaches are illustrated along with the predicted distributions of CO2 breakthrough capacity for a scaled-up process. © 2016 American Institute of Chemical Engineers AIChE J, 62: 3352–3368, 2016