Modelling Turbulent Non-Premixed Combustion in Industrial Furnaces

Abstract

Measuring the temperature distribution inside the rotary kilns, using thermocouples for instance has proven to be difficult due to the harsh operating conditions of the kiln. Numerical modelling of turbulent combustion and the associated physical phenomenon thus proves to be an indispensable tool towards predicting the kiln operating conditions. The purpose of the present work was to make a step towards modelling the cement rotary kiln used by Almatis B.V. in Rotterdam for the production of calcium-aluminate cement. The detailed mathematical model of the rotary kiln would be developed using the open source CFD toolbox OpenFOAM. The main advantage of OpenFOAM is that, contrary to most of the commercial CFD software, it is license fee free and allows access to the source code, which was also the motivation behind this work. To accurately model the Almatis kiln the following important phenomenon have to be taken into account: turbulent non-premixed combustion of hydrocarbon gases in the burner, radiative heat transfer distribution in the kiln and, the conjugate heat transfer through the furnace walls. In the present work the new solver implemented in OpenFOAM for turbulent combustion and radiation modelling was validated using the benchmark Sandia Flame D test case. There was good agreement seen between the results from simulations and experimental data for the Sandia Flame D test case indicating the adequacy and accuracy of the implemented transient solver and its readiness for further combustion application development. Due to the very complex geometry of the Almatis Kiln the relatively simple geometry of the Burner Flow Reactor (BFR) was considered for further simulations. The simulation results obtained for the Burner Flow Reactor (BFR) were compared with the commercial package ANSYS Fluent for consistency. The OpenFOAM toolbox was evaluated in two stages of increasing complexity: isothermal(cold) flow simulation and non-premixed gas combustion simulation using a turbulent incompressible flow solver. The cold flow comparison gave almost identical results for both OpenFOAM and ANSYS Fluent. However the reacting flow results showed varying agreement with ANSYS Fluent. The mass fraction of species showed good agreement but the temperature profile showed some deviations. With more stringent global NOx emission standards, predicting NOx formation in industrial furnaces is now a priority. The CFD modelling of pollutant NOx formation was considered in the present work. A new solver in OpenFOAM was developed for thermal NO prediction. The solver was validated with the ANSYS Fluent NOx post-processing utility using the Burner Flow Reactor geometry. The effectiveness of NOx reduction mechanisms including the variation of air to fuel equivalence ratio and flue gas re-circulation (FGR) was demonstrated using the Burner Flow Reactor test case. From this study it was concluded that OpenFOAM is a promising toolbox for modelling turbulent combustion and can be used for predicting the operating conditions of complex industrial furnaces. The current bottleneck identified with OpenFOAM is the very high computational cost of the implemented transient solver for turbulent combustion and radiation modelling. The computational cost of the transient solver far exceeds that of the steady state solvers available in commercial packages for example ANSYS Fluent. Therefore, to simulate very large scale industrial furnaces such as the Almatis Kiln in realizable time the implementation of a steady state solver for turbulent combustion applications in OpenFOAM is indispensable. It would also be essential to include the accompanying phenomenon of conjugate heat transfer into the solver. These can be accomplished as a part of the future work.