Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Chemical and Biomedical Engineering

Committee Chair

Debangsu Bhattacharyya

Committee Co-Chair

David S Mebane

Committee Member

David C Miller

Committee Member

Richard Turton

Committee Member

Stephen Zitney


To accelerate the development and commercial deployment of CO2 capture technologies, computational tools and models are being developed under the auspices of the U.S. Department of Energy's Carbon Capture Simulation Initiative (CCSI). The CCSI process modeling team was tasked with developing a "gold-standard" model that will serve as a definitive reference for benchmarking the performance of solvent-based CO2 capture systems under steady-state and dynamic conditions over a large operating-range. The main three areas that this work focused on are: development of the hydrodynamic and mass transfer submodels for a monoethanolamine (MEA) solvent system, uncertainty quantification of these submodels, development of a dynamic model for this system, and development of a dynamic design of experiment methodology for model validation and parameter estimation of this system.;For the gold-standard model, it was desired that the pressure drop and holdup models must be applicable over a wide range of operating conditions. In this work, a large range of liquid and gas flowrates, and wide range of viscosity and density for the liquid phase are considered and an optimal model is developed. The pressure drop and holdup models are also evaluated with data from numerous process scales.;Typically the mass transfer models and their parameters such as the liquid and gas-side mass transfer coefficients, diffusivity, and interfacial area are regressed using the data obtained from different experimental set-ups and scales, often in a sequential and sub-optimal way. In this work, a novel methodology is developed where parameters of the mass transfer models are simultaneously regressed by using the data from the wetted wall column, and packed towers, simultaneously. It is observed that the technique helps to improve the predictive capability of the process model.;Uncertainty in process models and their parameters are unavoidable. A Bayesian uncertainty quantification technique is applied for the first time to quantify the parametric uncertainty of the hydraulic and mass transfer models.;Dynamic models of CO2 capture solvent systems are very few in the existing literature. Model validation with the dynamic data from pilot plant has been scarcely reported. In this project, dynamic models are developed in Aspen Plus DynamicsRTM. Approximate pseudo random binary sequences are designed for the input signals and applied to the National Carbon Capture Center (NCCC) pilot plant during the 2014 MEA campaign. The pilot plant data were found to be noisy, did not satisfy mass and energy balances. In addition, some key variables were not measured. Preprocessing of the data followed by solution of a dynamic data reconciliation problem showed that the model could predict the transient response reasonably well.;For the first time, a dynamic design of experiments (DDoE) is developed for solvent-based CO2 capture processes using pseudo-random binary sequence and Schroeder-phased input techniques. The design ensured plant friendliness and could be successfully implemented in NCC during the 2017 campaign. The transient data are used to solve a dynamic data reconciliation and parameter estimation problem. Due to the computational expense and large dimensionality of the underlying problem, only the parameters corresponding to the holdup model could be estimated. It is observed that the holdup parameters could be optimally estimated by using the dynamic data collected over only a day. The parameters are slightly superior to those that have been regressed by using a large amount of the steady-state data collected over weeks. The techniques shows promise for the model development and parameter estimation by using the dynamic data that can be collected very quickly as opposed to the traditionally used steady-state data that take months thereby saving considerable resources.