A double contra-rotating axial swirler was investigated at 0.38 equivalence ratio and 600K inlet temperature at 1 bar, both experimentally and using CFD. Natural gas fuel was injected from between the two contra-rotating flows. The experimental results showed that whilst the NOx emissions were relatively high, the system has excellent flame stability characteristics meaning the flame could be maintained at very low fuel-air ratios, down to an equivalence ratio of around 0.28. CFD predictions were carried out using the equilibrium pdf combustion model alongside the partially premixed combustion model using default model constants, and the NOx model was applied as a post-processing application. The results showed that whilst the location and shape of the flame and the combustor fuel-air distribution could not be precisely captured, the peak temperature, CO, UHC and NOx levels were correctly predicted. The influence of the turbulent Schmidt number, Sct, was therefore investigated in an attempt to improve the predictions. Lower values offered an improvement in the fuel-air mixing predictions, but not in the combustion predictions, where the peak flame temperature could only be predicted correctly when using the default value of Sct = 0.85. This indicates that the use of a universal turbulent Schmidt number for both the fuel-air mixing and combustion models may not be suitable for the present swirler and combustor configuration, therefore future work will look to use a separate value of Sct for the mixture fraction than for the pdf combustion models.

A radial swirl low NOx combustor was investigated using CFD at 0.5 equivalence ratio and 600K inlet temperature at 1 bar. The equilibrium pdf combustion model using 16 species chemistry was shown to give a slightly improved prediction for the temperature distribution, but a poorer prediction for the CO distribution, over similar work using the flamelet model with 53 species chemistry. The NOx model was applied as a post-processing application and was shown to give a vastly superior result for the equilibrium pdf model using 16 species chemistry over the flamelet model using 53 species chemistry. Various NOx model configurations were tested and it was shown that only the mixture fraction based turbulence chemistry interaction model was able to provide a good result and all other models drastically under-predicted the peak nitric oxide levels. Combustion model predicted O modeling with excluded and combustion model predicted OH modeling were both shown to give a good match against measurements for the peak nitric oxide levels within the combustor when using mixture fraction based turbulence chemistry interaction.

A radial swirl DLE combustion system was investigated for its gaseous fuel-air mixing performance using various different RANS turbulence models. Two different configurations were investigated; vane passage fuel injection and fuel injection from the wall of an outlet throat directly into the shear layer. The results showed that for vane passage fuel injection, only the standard k-ε and standard Spalart-Allmaras models were able to provide a reasonable prediction for the combustor fuel-air distribution, out of all the RANS models and their variants available in Fluent v6.3, although both models predicted that the fuel and air were better mixed than in the measurements. For outlet throat wall fuel injection no models were able to provide a reasonable prediction. This same issue is also reported by several other researchers and represents a serious problem area in combustion modeling for low NOx applications. Improved fuel jet penetration was achieved by using an extremely low value of 0.1 for the turbulent Schmidt number, therefore future work will concentrate on using a localized value of Sc in the vicinity of the fuel injection hole.

A radial swirler with vane passage fuel injection using a radial fuel spoke with one fuel hole per passage was investigated using CFD at 0.5 equivalence ratio and 600K inlet temperature at 1 bar. Experimental measurements of the internal flame composition from water cooled gas sample probes were the experimental results used for comparison. Three combustion models were compared: flamelet with two difference kinetic schemes; PDF transport with two step chemistry and finite rate eddy dissipation model. Both models consistently underpredicted the turbulent flame thickness to 90% heat release by a factor of about 2. The PDF model with postprocessing NOx predictions over estimated the NOx emissions considerably and the best model was the flamelet model with full chemistry. The under prediction of the turbulent reaction zone thickness was concluded to be due to inadequate modelling of strained flame quenching for very lean flames with large laminar flame thickness and very low burning velocities. This flamelet model was applied to predict the influence of the radial swirler outlet geometry on the flame development, fuel and air mixing and NOx emissions. A dump expansion from the radial swirler outlet was compared with the addition of a shroud at the outlet and with the addition of a 60mm long outlet throat. The shroud was shown to increase the peak turbulence and confine it very close to the shroud lip. This improved the fuel and air mixing and lowered the predicted NOx from 2.7ppm to 1.2ppm with the shrouded swirler and 0.3ppm with the 60mm outlet throat and mixing length.

Well mixed low NOx gas turbines are limited, in the operational range of the low NOx mode, by the weak extinction and CO limits of the flame stabiliser used. The operational range of the combustor in the <10ppm low NOx mode is set by the range of equivalence ratios over which ultra low NOx without acoustic resonance can be achieved. This paper reviews the available data on weak extinction in well mixed low NOx combustion systems and presents some new data. Atmospheric pressure weak extinction data is shown to be similar to weak extinction at pressure for similar stabiliser designs and reference velocities. For low NOx gas turbine combustion it is demonstrated that all the best weak extinctions are identical to the lean flammability limit for laminar flames. Weak extinction is where the flow velocity exceeds the turbulent burning velocity and data on weak extinction is used as a measure of the mean turbulent burning velocity and shown to correlate with turbulent burning velocity data and theories. Methods of predicting the peak turbulence generated downstream of a flame stabiliser are outlined, based on grid plate measurements of turbulence and pressure loss. It is shown that a wide range of premixed flame stabilisers including swirling and non-swirling flame stabilisers have a weak extinction that can be predicted using this method.

A radial swirler was investigated with central fuel injection in the radial outward direction. 8 radial swirl vanes were used with 8 central radial fuel injector holes. The experimental results showed that central radial outward fuel injection aim had improved flame stability relative to vane passage fuel injection without the large NOx penalty associated with conventional fuel injection. The radially outward fuel jets impinged on the radially inward air from the vane passages and relatively good mixing occurred, but with sufficient unmixedness to stabilise the flame. CFD investigations were carried out to investigate this partial premixed and partial diffusion combustion mode which is commonly used for a pilot fuel location in lean low Nox combustion. RANS CFD predictions were undertaken with three combustion models: flamelet, pdf transport and eddy dissipation. The flamelet model was far superior both in the prediction of the measured quantitative values and in their distribution. The eddy dissipation model grossly over predicted the peak temperature and NOx. However, all the RANS models did not predict the measured direction of flame development, which was more in the axial direction on the inside of the shear layers. The RANS models predicted that the flame development would follow the shear layer. In an attempt to better predict the direction of flame development an LES model has been used. This indicates a better prediction of the direction of fuel and air mixing.

Radial swirlers with vane passage natural gas injection, similar to those used in some industrial low NOx gas turbines, were investigated for their flame structure both experimentally and using CFD. The radial swirler NOx emissions at 600K and 1 atmosphere pressure were shown to be 3-4 ppm at 15% oxygen at 1800K and 1-2 ppm at 1700K. These levels were similar to the best published low NOx emissions using any flame stabilizer design. A flame at empty set=0.5 and 600K air temperature was investigated for its structure using a 10mm OD water cooled gas sample probe with a I mm gas sample inlet on the upstream side of the probe. This showed that the mixing in the vane passage and outlet duct was very good. The maximum unmixedness at the first traverse location, 10mm downstream of the dump expansion zone, was 20% of the mean and the unmixedness was less than 5% within 30mm from the dump expansion. The flame structure was shown to involve a thick turbulence reaction zone of about 100mm thickness to the 90% heat release point. The CFD predictions were made using the RSM and k-epsilon turbulence models and the flamelet combustion model with a strain rate library. The isothermal aerodynamics predictions were in good agreement with others for similar geometries. There was an inner and outer recirculation zone with a swirling shear layer between. The peak turbulent kinetic energy was predicted to be on the inside of the shear layer. The experimental results showed that the flame developed in this region of high turbulence and low axial velocities. The flamelet model was less successful at predicting the flame development. The NOx results were predicted to be 2ppm less than the experimental results, due to the shorter predicted heat release region with associated lower prompt NOx.