Numerical and Experimental Investigations of the Siemens SGT-800 Burner Fitted to a Water Rig

Abstract

DLE (Dry Low Emission) technology is widely used in land based gas turbines due to the increasing demands on low NOx levels. One of the key aspects in DLE combustion is achieving a good fuel and air mixing where the desired flame temperature is achieved without too high levels of combustion instabilities. To experimentally study fuel and air mixing it is convenient to use water along with a tracer instead of air and fuel. In this study fuel and air mixing and flow field inside an industrial gas turbine burner fitted to a water rig has been studied experimentally and numerically. The Reynolds number is approximately 75000 and the amount of fuel tracer is scaled to represent real engine conditions. The fuel concentration in the rig is experimentally visualized using a fluorescing dye in the water passing through the fuel system of the burner and recorded using a laser along with a CCD (Charge Couple Device) camera. The flow and concentration field in the burner is numerically studied using both the scale resolving SAS (Scale Adaptive Simulation) method and the LES (Large Eddy Simulation) method as well as using a traditional two equation URANS (Unsteady Reynolds Average Navier Stokes) approach. The aim of this study is to explore the differences and similarities between the URANS, SAS and LES models when applied to industrial geometries as well as their capabilities to accurately predict relevant features of an industrial burner such as concentration and velocity profiles. Both steady and unsteady RANS along with a standard two equation turbulence model fail to accurately predict the concentration field within the burner, instead they predict a concentration field with too sharp gradients, regions with almost no fuel tracer as well as regions with far too high concentration of the fuel tracer. The SAS and LES approach both predict a more smooth time averaged concentration field with the main difference that the tracer profile predicted by the LES has smoother gradients as compared to the tracer profile predicted by the SAS. The concentration predictions by the SAS model is in reasonable agreement with the measured concentration fields while the agreement for the LES model is excellent. The LES shows stronger fluctuations in velocity over time as compared to both URANS and SAS which is due to the reduced amounts of eddy viscosity in the LES model as compared to both URANS and SAS. This study shows that numerical methods are capable of predicting both velocity and concentration in a gas turbine burner. It is clear that both time and scale resolved methods are required to accurately capture the flow features of this and probably most industrial DLE gas turbine burners.