CFD & AMP Center
Department of Mechanical & Aerospace Engineering
West Virginia University
E: ismail.celik@mail.wvu.edu

Prediction of Gas Turbine Combustion Emissions using RANS/EDC and RANS/PDF methods

Sponsors
NETL/DOE
 
Researchers
Dr. Ismail B. Celik
Dr. Jaggu Nanduri
Don Parsons
 
Collaborators
Dr. Payman Givi (University of Pittsburgh)
 

It is well known that RANS (Reynolds Averaged Turbulence Models) and Turbulent Chemistry Models (i.e. combustion models) including scalar transport, and reaction mechanisms are the two primary factors influencing the uncertainty in predicting emissions from turbulent combustion in general, and gas-turbine combustion (GTC) in particular. This project aims to assess current state-of-the art turbulence models and combustion models with respect to their performance in gas-turbine combustion emissions predictions for various practical fuels and fuel blends that are currently being used in microturbine applications. Reliable prediction of emissions for various fuel blends will allow the microturbine developers to operate the turbine using various fuels and fuel blends without having to perform costly validation studies. Utilization of CFD for combustor design could lower the design costs to a fraction of experimental costs. In fact, RANS and PDF (Probability Density Functions) based turbulent combustion models are already used in the industry as a combustor design tool. However the cost of computing and the reliability of the emissions predictions prevent most developers from using detailed reaction mechanisms and NOx mechanisms in the design process. Despite all of the recent developments in both RANS and Large Eddy Simulation (LES) for prediction of turbulent combustion, most current commercial codes are based on mostly out-dated (and in some cases very crude) methodologies. In recent years the limitations and drawbacks of classical “popular” methods such as the eddy break up, Bousinnesq-based approximations (including the k-e), etc. in RANS and Smagorinsky, flamelet, etc. in LES have been recognized. However, these methods are still predominant in almost all the current commercial (and many research) codes. With this assessment, we aim at demonstrating the actual capability of selected RANS models some of which could be used in hybrid RANS/LES applications. The outcome of this study will also serve as a basis for further studies involving LES.

SimVal is an optically accessible research combustor specially designed for CFD model validation studies developed by NETL (National Energy Technology Laboratories). The geometry of the swirl stabilized combustor and the swirler is shown in Figure 1. This research combustor is being used to develop an experimental database of emissions as a function of fuel composition. Validation of the current CFD study will be performed against this database.


Figure 1 - SimVal Combustor Schematic* and the CFD grids used
* Starkey, P., Sidwell, T., and Ontko, J., Proceedings of the Combustion Institute, Vol. 31, pp xxx-xxx (In Press), (2006)

Combustion calculations were conducted on WVU SimVal grid (Figure 1 above) using ARM9 [1] and ARM19 [2] mechanisms using the RNG k-e and EDC combustion model. Calculations are performed using the commercial CFD software FLUENT. Results of the ARM9 mechanism for various lean equivalence ratios for pure methane and 20% hydrogen + 80% methane blends are shown below.




Figure 2 - Comparision of NO mass fractions (lines in ppm) and temperature (filled contours) for 100% CH4 flame (left) and 20% H2 + 80% CH4 flame (right) at equivalence ratios 0.46 (top), 0.56 (middle) and 0.60 (bottom)

The results show qualitative agreement with the experimental SimVal results. The reduction in the NO due to hydrogen addition is also seen in the experiments. Further calculations are being conducted to improve the boundary conditions to match the experiments. Reduced mechanisms with higher numberer of species and reactions are also being considered. PITT is considering RANS/PDF based turbulent combustion models to simulate similar flames.