Simulation studies on a non-isothermal, non-adiabatic, fixed-bed reactor

A gas-solid catalytic fixed-bed reactor was studied via numerical simulations, using both pseudo-homogeneous and rigorous heterogeneous transient models with the choice of two different kinetic models for the partial oxidation of ortho-xylene, in the presence of vanadium pentoxide (anatase) catalysts. The research project is positioned within the context of computer-aided process engineering, with improvement of existing designs and operating parameters as its primary goal. The purpose of this study is to provide general recommendations for selecting models and numerical methods for constant wall-cooled fixed-bed reactor simulations, as well as a methodology for optimizing the operating parameters.;The research project compared kinetic models that treat the ortho-xylene reaction network at three different levels of sophistication. The study also compared the modeling of the gas-solid reactions in a fixed-bed catalytic reactor using the simplified pseudo-homogeneous approach and the rigorous heterogeneous approach.;The dissertation presents examples of and results on the following computational topics: two-step and one-step non-linear optimization methods for parameter estimation; parametric sensitivity analysis using the direct differential method; numerical solution techniques for stiff differential (partial and ordinary) algebraic equation systems; and the tools for parallelization of reactor simulation codes for execution on high-performance computers.;The correct choice of the kinetic model plays the most critical role in the realistic outcome of the reactor simulation. In particular, using a simplified kinetic model may result in either conservative or erroneous conclusions in parametric sensitivity. The choice of different reactor models also becomes important if transient simulations are required. For parameter estimation, the MLVI (two-step) method is less efficient than the variational analysis based (one-step) method, but it is less sensitive to the consistency of initial values and easier to implement. The variational analysis based method, on the other hand, was shown to be a valuable non-linear optimization method for continuous variables. Reactor parametric sensitivity analysis is an important step in the search for favorable operating conditions; both geometry- and sensitivity-based criteria were successfully applied and provided similar results.;The SDIRKDAE solver (developed locally) was shown to be efficient for solving transient stiff problems, especially for near-square domain, spatially discretized PDE equations. A linear, approximate analytical Jacobian proposed for reaction-transport systems was shown to be effective for improving the small-step transient simulation speed. Finally, it was demonstrated that parallel computation using Cactus and PETSc significantly improves the temporal performance of high-resolution dynamic reactor simulations.