(162f) Numerical Evolution of Model and Parametric Sensitivities in Turbulent Reacting Flow Simulations

Computational modeling of turbulent reacting flows has become an invaluable tool in the design and scale-up of chemical reactors. In spite of vast advances in the understanding of basic flow and reaction physics, computational models still lack broad predictive capability. A key source of error is the modeling coefficients used in the description of the different physical processes such as turbulence or chemical reactions. These coefficients are often derived empirically, without accounting for the interactions amongst the processes. Hence, the simulations are highly sensitive to the values used for these coefficients. Developing better models requires that the impact of coefficient variations on simulation results is understood. In order to estimate this sensitivity, we propose a novel sensitivity evolution methodology. We develop a non-intrusive technique, where partial differential equations governing the evolution of the sensitivity of the turbulent flow field to the model coefficients are solved along with the modeled flow equations. This methodology is applied to the large eddy simulation of gas-phase aerosol synthesis process. Here, the uncertainties in the model coefficients for aerosol nucleation, aggregation, and growth are studied. In addition, the propagation of errors due to uncertain boundary conditions is also studied. Interpretation of the sensitivity results and their use in model development will be discussed.