Experimental and computational study of hybrid diffusers for gas turbine combustors
2010-01-25T12:37:05Z (GMT) by
Diffusers are essential in gas turbine combustors, decelerating the compressor efflux prior to the combustion chamber to reduce total pressure losses. Modern, low emission, radially staged combustors require even more diffusion due to the increased flame tube depth of this type of combustor. Furthermore, these high rates of deceleration are accompanied by large adverse pressure gradients and an associated risk of flow separation. Previous studies have shown that hybrid diffusers can achieve high rates of efficient diffusion in far shorter lengths than conventional faired diffusers or dump diffuser systems, representing a potential performance gain and weight saving. Hybrid diffusers consist of a wide angle diffuser immediately downstream of a sudden expansion, with flow separation prevented by bleeding off a small amount of the mainstream flow. However, previous studies have not provided a conclusive understanding of the associated flow mechanisms leading to hybrid diffusers currently being considered high risk. Additionally definitive data does not exist on the influence of bleed gap geometry and therefore hybrid diffusers cannot, currently, be optimised for use in a modern gas turbine. Further issues also not addressed by earlier studies, but concerning the use of hybrid diffuser in gas turbine combustors, are the effect of representative inlet conditions incorporating vane wakes at diffuser inlet, the quality of the bleed air and its potential for use for component cooling, the effect of radial struts within a hybrid diffuser and the quality of the flow delivered to the combustor feed annuli (total pressure losses). Therefore, a predominately experimental study, coupled with CFD predictions, was undertaken to investigate the controlling flow mechanisms of hybrid diffusers and address the questions necessary to evaluate the suitability of hybrid diffusers for use in modern, low emission, radially staged combustion systems. An existing isothermal test facility was used comprising a fully annular, staged combustor downstream of a single stage axial compressor incorporating engine representative outlet guide vanes. Initial experimental work led to rig modifications which allowed a range of hybrid diffusers to be studied. To act as a benchmark the performance of a conventional single-passage, dump diffuser system was first studied. A hybrid diffuser demonstrated a 53% increase in area ratio within the same axial length as the conventional diffuser. Results showed that this hybrid diffuser achieved a 13% increase in static pressure recovery which, in turn, improved the feed to the combustor feed annuli and decreased total pressure loses by 25%. Notably this brought the annulus losses within accepted target values; something the conventional diffuser system was unable to do. Additionally, it was clearly shown, in contradiction to previous studies, that bleeding air via a vortex chamber was not necessary. Bleeding air via a simple duct arrangement achieved the same results without altering the governing flow mechanisms. To provide a better understanding of these flow mechanisms, a computational investigation was also undertaken. A commercial CFD code, Fluent, was used to solve the Reynolds averaged Navier-Stokes equations for an incompressible flow regime, employing a blended second order upwind/central differencing scheme and the SIMPLE pressure correction algorithm. The turbulence was modelled using the k-ε model in conjunction with a standard wall function. Several generic two-dimensional hybrid diffusers were studied in order to reveal the controlling flow mechanisms and enable optimisation of the bleed gap geometry. Importantly, this revealed that many features previously thought to contribute to the flow mechanisms were, in fact, unnecessary. A detailed examination of the flow field, including an analysis of the terms within the momentum equation, demonstrated that the controlling flow mechanisms were not simply a boundary layer bleed but involve a much more complex interaction between the accelerating bleed flow and the diffusing mainstream flow. Firstly, momentum is transferred from the accelerating bleed flow to the diffusing mainstream flow, enabling a fresh boundary layer to be formed on the diffuser wall which is sufficiently energetic to overcome the high rates of diffusion and high adverse pressure gradient. Secondly, the radial pressure gradient created by the bleed causes deflection of the mainstream flow which also transports higher momentum fluid into the boundary layer. Understanding this resulted in a greatly simplified design for the hybrid diffuser not only potentially reducing weight but also reducing bleed flow total pressure losses. Predictions for a three-dimensional representation of the experimental facility displayed many similarities in the flow field and similar performance trends to the experimental data. However, predicted values of total pressure loss and static pressure recovery differed from experimental data and it was thought that this was due to an incomplete description of the turbulence (k and ε) at inlet and/or known problems the k-ε turbulence model has with predicting some unconfined flows. Nonetheless, three-dimensional predictions revealed an interaction between the OGV wake fluid and bleed flow causing localised, but small, modification of the flow mechanisms. Furthermore, it was shown that without the levels of turbulence produced downstream of an axial compressor the hybrid diffuser under study would, in fact, stall. Overall, experimental and computational results obtained in the current research suggest that the performance of hybrid diffusers is more than satisfactory for use within lowemission, staged, gas turbine combustion systems. An understanding of the governing flow mechanisms and the effect of features such as OGV wakes or radial struts has lead to a more practical design of hybrid diffuser, simplifying the geometry and reducing bleed flow total pressure losses (increasing the possibility of this air being used for component cooling).