Application of a Validation/Uncertainty Quantification (VUQ) Methodology at Two Scales: From Modeling of Char Oxidation to Simulation of a 1.5MW Coal-Fired Furnace

摘要

This dissertation presents two validation/uncertainty quantification (VUQ) studies, one on the 1.5 MWth coal fired furnace (L1500) and the other on the Reacting Particle and Boundary Layer (RPBL) char oxidation model. A six-step methodology is used in both cases.;In Chapters 2 and 3, the VUQ for the L1500 furnace is presented; the quantities of interest (QOIs) are the heat removal by the cooling tubes and the wall temperature. In Chapter 2, the Arches simulation of the L1500 base case is described in detail. From the simulation, two models that impact the QOIs are selected for further analysis, the ash deposition model and the char oxidation model. An input and uncertainty (I/U) map is created from the parameters in these models and a sensitivity analysis is performed with the five parameters that have the greatest impact on the QOIs. From the sensitivity analysis, two parameters (thermal conductivity of the deposit and wall emissivity) are chosen for the next steps in the VUQ cycle. In Chapter 3, an updated version of the I/U map with two additional parameters, the coal feed rate and the swirl factor (this factor varies the tangential component of the velocity), is presented. The thermal conductivity of the deposit and wall emissivity are combined into one parameter, the effective thermal conductivity. These three active parameters are then used in the consistency analysis. The experimental uncertainty of the QOIs is estimated by adding the sampling and the systematic errors. Data collection for the simulation is done with 34 cases obtained by varying the three active parameters. For each experimental QOI, a Gaussian process (GP) surrogate model is built from the set of simulation data. The consistency analysis is performed with the GP surrogate models and the QOIs with their estimated uncertainties; consistency is achieved. Chapter 3 concludes with recommendations for reducing the uncertainty in the experimental measurements and a review of model assumptions.;In Chapters 4 and 5, the VUQ for the RPBL model is presented; the QOIs are the particle temperature and velocity. In Chapter 4, the RPBL model formulation and the associated I/U map are given. One case is presented and explained in detail. A sensitivity analysis with nine parameters from the I/U map is performed, and five parameters (dp,&phis;initial, Yc, epsilon p, and rinf/rp) are selected for the next steps. In Chapter 5, the priorities of the parameters in the I/U map are updated using the results from sensitivity analysis. To compute the experimental uncertainty associated with the QOIs, the main contribution is assumed to be the sampling error. The RPBL model is run with the five parameters to produce simulation data. A polynomial chaos (PC) surrogate model is built for the set of simulation data corresponding to each experimental QOI. The consistency analysis is performed with the PC surrogate models and the QOIs with their estimated uncertainties. Consistency is found for three different types of char and six different particle size ranges in two O2 environments, each with 6-14 QOIs. To conclude, Chapter 5 reviews the experimental measurements, analyzes what is learned about the parameters in the consistency analysis, and revisits the assumptions made in the RPBL formulation.