Kiln aerodynamics: influence from turbulence models and boundary conditions

Abstract: LKAB (Luossavaara-Kiirunavaara AB) is an international high-tech minerals
group that produces iron ore products for the steel industry. LKAB's main
product is iron ore pellets. In the pelletising process nitrogen oxides,
among other emissions, are produced. Stricter environmental laws results in
higher demands on the companies and their environmental impact. LKAB is
therefore running several development projects with the objective to reduce
NOx emissions from their pellet production plants.
In this project the focus has been on the grate-kiln process and especially
on the rotary kiln. A down-scaled model of the real kiln has been created
and with help of Computational Fluid Dynamics (CFD) the flow field has been
investigated. The objective was to get a deeper understanding of the flow
field in order to decide if changes of the geometry up-stream the burner
can change the flow field, improve the combustion and reduce the
environmental harmful NOx emissions.
The commercial code ANSYS CFX was used and different boundary conditions
and turbulence models and their impact on the flow field were studied. The
turbulence models used were the SST-model, which is a two-equation model,
and the BSL-model, which is a Reynolds stress model. The most prominent
result was the differences between the two turbulence models. The
turbulence model's ability to reveal secondary flow affected the flow field
pattern and the result with plug profiles as inlet boundary conditions
deviated from the one where velocity profiles were used.
One conclusion was that a more advanced turbulence model can be used when
simulating a down-scaled model like this one. If the secondary flows are of
interest, two-equation models should not be used. The two-equation model
showed trouble with convergence in certain cases, especially when velocity
profiles were used as inlet boundary conditions. The more advanced Reynolds
stress model created stable simulations with good convergence in every case
at no noticeable increased run time.