Overview
The SAGE6 demonstration project aimed to develop and mature a lean burn combustion system. An essential enabler to development of such technology was an accurate and reliable computational tool for prediction of emissions. Lean burn provides significant benefits in terms of NOx emissions. However, the emissions of CO, UHC and soot limit the operation of the combustor at different conditions. Reliable predictions of emission trends led to optimised combustor designs in a cost effective way. Today’s capabilities, however, are still inadequate to produce accurate and reliable predictions in direct support of lean burn system design. The DREAMCODE project aimed to develop and improve computational methods that can be used in the design process of low emission combustors. Improved models and methods were developed to predict emissions accurately and reliably. To that end, the following essential elements of a CFD combustion emission tool were considered:
- Detailed chemistry models for jet fuel surrogates are necessary to describe the complicated chemical processes of fuel oxidation and emission formation in the gas phase.
- Soot models are indispensable to describe the complex physical and chemical phenomena of soot particle formation.
- Chemistry reduction methods are inevitable to reduce the computational cost of the complex chemistry model for application in CFD codes.
- Spray break-up models are necessary to model the liquid fuel break-up, which has a dramatic effect on emissions.
- Turbulence-chemistry interaction models have to account for the effects that occur on length scales which cannot be resolved by the computational mesh.
These 5 models were improved and integrated in a CFD code for the validation on real aero engine gas turbine combustors.
Funding
Results
Executive Summary:
Project DREAMCODE improved computational methods developed to predict combustion emissions accurately and reliably. DREAMCODE was part of the SAGE6 demonstration project, which aimed to develop and mature a lean burn combustion system. Lean burn provides significant benefits in terms of NOx emissions. However, the emissions of CO, UHC and soot limit the operation of the combustor at different conditions. Reliable predictions of emission trends led to optimised combustor designs in a cost effective way. Existing capabilities, however, were inadequate to produce accurate and reliable predictions in direct support of lean burn system design. To develop computational methods that can be used in the design process of low emission combustors, the following essential elements of a CFD combustion emission tool were improved.
First, detailed comprehensive reaction mechanisms for jet fuel surrogates were necessary to describe the complicated chemical processes of fuel oxidation and emission formation. In DREAMCODE an extensive mechanism for jet fuel surrogates was step by step updated, refined and validated. The new mechanism was found to work well for a wide range of test cases with significant improvements in predictions of NOx and soot precursors.
Second, soot models were indispensable to describe the complex physical and chemical phenomena of soot particle formation. In DREAMCODE the modelling of soot formation was advanced by the development of a new multivariate soot model and improved statistical models. Model uncertainties have been analysed systematically and carefully validated. Satisfactory predictions of soot volume fractions and particle size distributions were obtained.
Third, reduced chemistry method were developed to include detailed chemistry models in CFD calculations at affordable computational costs. More specifically, the Flamelet-Generated Manifold (FGM) reduction method was coupled with the soot models and extended to improve the prediction of combustor emissions. In addition, the Rate Controlled Constrained Equilibrium (RCCE) method was employed to reduce detailed mechanisms for kerosene. By combining the RCCE with artificial neural networks a speed-up of one to two orders of magnitude was obtained.
Fourth, accurate and efficient spray break-up models were developed to predict the fuel distribution in the combustor. To that end, atomisation experiments were analysed with advanced analysis tools to extract the characteristics of the atomization process. In addition, detailed Smoothed Particle Hydrodynamics (SPH) simulations were performed to study the fuel break up in pre-filming air-blast atomizers. Based on these new insights a phenomenological spray model for predicting the droplet size distribution was developed and successfully implemented into a CFD code.
Finally, the various models were integrated in a CFD code for the modelling of turbulent combustion in gas-turbine combustors. Large-Eddy Simulations (LES) of various test cases were performed to assess the accuracy of the models. With the integration of these improved models, a much more reliable CFD tool has been developed that will guide the further design of lean burn combustion systems.