Computational Fluid Dynamics Unstructured Mesh Optimization for the Siemens 4th Generation DLE Burner

Every computational fluid dynamics engineer deals with a never ending story – limitedcomputer resources. In computational fluid dynamics there is practically never enoughcomputer power. Limited computer resources lead to long calculation times which result inhigh costs and one of the main reasons is...

Full description

Bibliographic Details
Main Author: Koren, Dejan
Format: Others
Language:English
Published: KTH, Mekanik 2015
Subjects:
CFD
CFX
Online Access:http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-178034
Description
Summary:Every computational fluid dynamics engineer deals with a never ending story – limitedcomputer resources. In computational fluid dynamics there is practically never enoughcomputer power. Limited computer resources lead to long calculation times which result inhigh costs and one of the main reasons is that large quantity of elements are needed in acomputational mesh in order to obtain accurate and reliable results.Although there exist established meshing approaches for the Siemens 4th generation DLEburner, mesh dependency has not been fully evaluated yet. The main goal of this work istherefore to better optimize accuracy versus cell count for this particular burner intended forsimulation of air/gas mixing where eddy-viscosity based turbulence models are employed.Ansys Fluent solver was used for all simulations in this work. For time effectivisationpurposes a 30° sector model of the burner was created and validated for the meshconvergence study. No steady state solutions were found for this case therefore timedependent simulations with time statistics sampling were employed. The mesh convergencestudy has shown that a coarse computational mesh in air casing of the burner does not affectflow conditions downstream where air/gas mixing process is taking place and that a majorpart of the combustion chamber is highly mesh independent. A large reduction of cell count inthose two parts is therefore allowed. On the other hand the RPL (Rich Pilot Lean) and thepilot burner turned out to be highly mesh density dependent. The RPL and the Pilot burnerneed to have significantly more refined mesh as it has been used so far with the establishedmeshing approaches. The mesh optimization has finally shown that at least as accurate resultsof air/gas mixing results may be obtained with 3x smaller cell count. Furthermore it has beenshown that significantly more accurate results may be obtained with 60% smaller cell count aswith the established meshing approaches.A short mesh study of the Siemens 3rd generation DLE burner in ignition stage of operationwas also performed in this work. This brief study has shown that the established meshingapproach for air/gas mixing purposes is sufficient for use with Ansys Fluent solver whilecertain differences were discovered when comparing the results obtained with Ansys Fluentagainst those obtained with Ansys CFX solver. Differences between Fluent and CFX solverwere briefly discussed in this work as identical simulation set up in both solvers producedslightly different results. Furthermore the obtained results suggest that Fluent solver is lessmesh dependent as CFX solver for this particular case.