In modern, low emission, gas turbine combustion systems the amount of air available for cooling of the flame tube liner is limited. This has led to the development of more complex cooling systems such as cooling tiles i.e. a double skin system, as opposed to the use of more conventional cooling slots i.e. a single skin system. An isothennal experimental facility has been constructed which can incorporate 10 times full size single and double skin (cooling tile) test specimens. The specimens can be tested with or without effusion cooling and measurements have been made to characterise the flow through each cooling system along with the velocity field and cooling effectiveness distributions that subsequently develop along the length of each test section. The velocity field of the coolant film has been defined using pneumatic probes, hot-wire anemometry and PIV instrumentation, whilst gas tracing technique is used to indicate (i) the adiabatic film cooling effectiveness and (ii) mixing of the coolant film with the mainstream flow. Tests have been undertaken both with a datum low turbulence mainstream flow passing over the test section, along with various configurations in which large magnitudes and scales of turbulence were present in the mainstream flow. These high turbulence test cases simulate some of the flow conditions found within a gas turbine combustor. Results are presented relating to a variety of operating conditions for both types of cooling system. The nominal operating condition for the double skin system was at a coolant to mainstream blowing ratio of approximately 1.0. At this condition, mixing of the mainstream and coolant film was relatively small with low mainstream turbulence. However, at high mainstream turbulence levels there was rapid penetration of the mainstream flow into the coolant film. This break up of the coolant film leads to a significant reduction in the cooling effectiveness. In addition to the time-averaged characteristics, the time dependent behaviour of the .:coolantfilm was. also investigated. In particular, unsteadiness associated with large scale structures in the mainstream flow was observed within the coolant film and adjacent to the tile surface. Relative to a double skin system the single skin geometry requires a higher coolant flow rate that, along with other geometrical changes, results in typically higher coolant to mainstream velocity ratios. At low mainstream turbulence levels this difference in velocity between the coolant and mainstream promotes the generation of turbulence and mixing between the streams so leading to some reduction in cooling effectiveness. However, this higher momentum coolant fluid is more resistant to high mainstream turbulence levels and scales so that the coolant film break up is not as significant under these conditions as that observed for the double skin system. For all the configurations tested the use of effusion cooling helped restore the coolant film along the rear of the test section. For the same total coolant flow, the minimum value of cooling effectiveness observed along the test section was increased relative to the no effusion case. In addition the effectiveness of the effusion patch depends on the amount of coolant injected and the axial location of the patch. The overall experimental data suggested the importance of the initial cooling film conditions together with better understanding of the possible mechanisms that results in the rapid cooling film break-up, such as high turbulence mainstream flow and scales, and this will lead to a more effective cooling system design. This experimental data is also thought to be ideal for the validation of numerical predictions.

Full-coverage film cooling is investigated both experimentally and numerically. First, surface measurements local of adiabatic film cooling effectiveness and heat transfer augmentation for four different arrays are described. Reported next is a comparison between two very common turbulence models, Realizable k-ε and SST k-ω, and their ability to predict local film cooling effectiveness throughout a full-coverage array. The objective of the experimental study is the quantification of local heat transfer augmentation and adiabatic film cooling effectiveness for four surfaces cooled by large, both in hole count and in non-dimensional spacing, arrays of film cooling holes. The four arrays are of two different hole-to-hole spacings (P/D = X/D = 14.5, 19.8) and two different hole inclination angles (α = 30◦ , 45◦ ), with cylindrical holes compounded relative to the flow (β = 45◦ ) and arranged in a staggered configuration. Arrays of up to 30 rows are tested so that the superposition effect of the coolant film can be studied. In addition, shortened arrays of up to 20 rows of coolant holes are also tested so that the decay of the coolant film following injection can be studied. Levels of laterally averaged effectiveness reach values as high as ¯η = 0.5, and are not yet at the asymptotic limit even after 20 − 30 rows of injection for all cases studied. Levels of heat transfer augmentation asymptotically approach values of h/h0 ≈ 1.35 rather quickly, iii only after 10 rows. It is conjectured that the heat transfer augmentation levels off very quickly due to the boundary layer reaching an equilibrium in which the perturbation from additional film rows has reached a balance with the damping effect resulting from viscosity. The levels of laterally averaged adiabatic film cooling effectiveness far exceeding ¯η = 0.5 are much higher than expected. The heat transfer augmentation levels off quickly as opposed to the film effectiveness which continues to rise (although asymptotically) at large row numbers. This ensures that an increased row count represents coolant well spent. The numerical predictions are carried out in order to test the ability of the two most common turbulence models to properly predict full coverage film cooling. The two models chosen, Realizable k − ε (RKE) and Shear Stress T ransport k − ω (SSTKW), are both two-equation models coupled with Reynolds Averaged governing equations which make several gross physical assumptions and require several empirical values. Hence, the models are not expected to provide perfect results. However, very good average values are seen to be obtained through these simple models. Using RKE in order to model full-coverage film cooling will yield results with 30% less error than selecting SSTKW.