Date of Graduation
Statler College of Engineering and Mineral Resources
Mechanical and Aerospace Engineering
An investigation was conducted to determine the feasibility of developing a reduced order model capable of accurately predicting the behavior of steam methane reforming. An emerging model reduction technique based on examining causal relationships was applied to the reaction network developed by Xu and Froment to eliminate unnecessary intermediate species. A dynamic discrepancy term was included in the reduced network to quantify the error incurred from the network reduction. This discrepancy is stochastic in nature, and a Markov Chain Monte Carlo (MCMC) sampling routine coupled with Bayesian statistical methods was used to calibrate the parameters of the discrepancy by comparison with simulated data provided by a more robust model of methane reforming.;An output distribution of discrepancy parameters was calibrated based on the transient response of a laboratory scale continuous stirred tank reactor (CSTR). Extrapolation to predictions of more complex plug flow reactor (PFR) models was also shown effective using the calculated distribution. Specifications regarding reactor geometry and operating conditions were taken from previously published studies. Traditional simplifying assumptions were specified to reduce the computation complexity of both the reactor simulation and calibration routine. Simulations were performed using a combination of MATLAB and C++ Matlab executable (MEX) files using high performance computing resources available through West Virginia University's Spruce Knob cluster.;Results of the calibration showed that the proposed modeling technique is able to reproduce the behavior of both the transient response of the single constant stirred tank reactor and the discretized plug flow reactor approximations. Convergence of the calibration routine was validated through statistical means. Additionally, computational times for both the robust model and the proposed reduced model are shown to be on the same order of magnitude. The combination of these findings verifies the ability of the proposed modeling technique to not only accurately predict the behavior of steam reforming but also indicates the potential for applying the proposed method for more complex simulations.
Ford, Evan D., "A Bayesian Approach to Reduced Order Modeling in Catalytic Steam Reforming" (2015). Graduate Theses, Dissertations, and Problem Reports. 5608.