DTI: 2010–11 Funded Proposal

Sean C. Garrick, Graham Candler, George Karypis

Turbulent Multi-Phase Flow Modeling, Data Mining, and Advanced Environmental Informatics for Mercury Removal from Coal Combustion Flue Gases

The EPA, and the states, have mandated a 50% to 90% reduction in the 140 tons of mercury released into atmosphere annually. There are several methods to control mercury emissions, including the injection of carbon particles into flue gas, where mercury vapor (a by-product of coal combustion) is adsorbed "inflight" by particles. We propose to develop a computational tool that utilizes high-performance computing, advanced modeling, and data mining that provides feature identification and extraction, to allow scientists and engineers to accurate predict flow dynamics and mercury capture in coal-fired powerplants. We will utilize numerical simulation to develop transferable data, new models and computational tools for the quantitative prediction of mercury adsorption gas-particle flows. By capturing the multi-scale interactions and the sorption characteristics of porous particles, numerical simulation with built-in feature identification & extraction can shed light on gas-particle mass transfer under turbulent flow conditions. To do this, we will utilize state-of-the-art turbulent flow code, coupled with multi-scale particle transport, and heterogenous Eulerian-Lagrangian data-mining (because a significant fraction of the computed variables are associated with the particles themselves). We will consider isothermal and non-isothermal processes, nano and micro-scale particles, and approaches that either project the variables to a background mesh and persist the information using the nodes of that mesh or write out the coherent structures by specifying the coordinates of all the particles involved in the mercury mass-transfer process. The PIs will leverage their experience in direct and large eddy simulation, multi-phase flows, data-mining, and dynamic feature identification and extraction to create tools (with a high-degree of fidelity to real world processes) that facilitate the control of fluid-chemical-particle dynamics for the removal of mercury generated during industrial processes.