Soot production in gas turbine combustors is not desirable since it is the major sourceof exhaust smoke emission and its thermal radiation to the combustor liner deteriorates the linerdurability. Soot formation involves comparatively slow chemistry and equilibrium can not beapplied to soot modelling in the combustor flow field.. The exact sooting process in thecombustor is poorly understood given both the complexity and the limited experimental dataavailable. The work reported in this thesis seeks to first develop in-situ techniques forretrieving spatially-resolved soot properties, mainly soot particle volume fraction, from withinthe combustor and also to apply the measured results to comparisons with predicted sootconcentrations.Two probing methods have been demonstrated which also incorporate a laser absorptiontechnique. The sight probe proves to be more reliable in the present measurements. Theevaluation of the physical probing techniques in sooty laboratory flames reveals that the flamestructure will not be substantially distorted by the probe. The disturbance caused by the probeis localised, a feature which is evident in the reported water flow visualization test. Thenecessary inert gas purge can be minimised to reduce the local aerodynamic perturbation. Themeasured soot volume fraction distributions are comparable with sooting levels reported inflame studies in the literature. The peak soot volume fractions are located off-axis,characteristic of the fuel atornization. The measurementsin the primary zone are restricted bythe multi-phase character of the flow, where soot absorption can not be readily discriminatedfrom fuel droplet scattering. Measurements are reported over a range of air-fuel ratios, inletpressures and temperatures.Time-averageds calard istributionsa t the nominald ilution sectionh ave beeno btainedin addition to the soot measuremenut sing probe sampling and standard gas analysis.Correlationso f carbond ioxide with mixture fraction reveala clear relationshipa t overall leanconditionsc onsistenwt ith widely usedm odelleda ssumptions.T here are less well-correlatedrelationshipsb etweent emperaturea ndm ixture fraction, possiblyd ue to the influenceo f scalarfluctuationsa nda lsoo f the scalard issipationr ate. Sootl oadingi n the presentf low conditionsis characteristicallylo w, basedo n the mixture fraction ands ootv olumef raction data. Thermalradiation in the visible spectrum shows a distinct narrow band spectra in addition to the sootcontinuum, which is believed to arise fromC2radical emission. The mean radiation intensities,predictedb y usingt he measuredte mperaturea nds ootc oncentrationre sults,a rei n generallo werthan the measured mean intensities. Temperature fluctuation levels may be particularlyinfluential in some of these calculations.Sootm odellingi n the combustohr asb eenu ndertakenb y applyinga n extendedla minarflamelet concept. The two-equations oot formation model has beenp rimarily developedo nlaminar flames. The comparisono f the computationa nd measuremenstu ggeststh at this sootmodel holds promise in the context of prediction in the combustor. In the absenceo f asatisfactoryt heoreticald escriptiono f the fuel-air burning in the combustor,w heret he liquidkerosinee mployedis replacedb y gaseoups ropane,t he computeds calarp rofiles are inconsistentin some importantr espectsw ith the measuredo nes. This exerts a major effect on the sootpredictioni n terms of the quantitatived etail in the computationw, hich is howeverc rucial forthe soot model development. The original flow field modelling needs to be improved for thepurpose of further soot model refinement.
展开▼
