Abstract
Efficient, accurate modelling of the internal air system is critical to the design of modern gas turbine engines and optimising the usage of cooling air allows for increased efficiency while maintaining design safety. Traditional 1D methods using empirical correlations are reliant on matching to engine data to produce reliable results. Coupled CFD/FE aero-thermomechanical models offer the promise of accurate prediction without being dependent on engine data. However, current coupled models, while giving accurate predictions during the stabilised regions of the flight cycles, show large errors during the transient phases. It is hypothesised that these transient errors are caused by the closed coupling loop of the labyrinth seal clearances determining the mass flows through the labyrinth seals, the mass flows through the labyrinth seals determining the metal temperatures, and the metal temperatures determining the labyrinth seal clearances. Consequently, the research in this thesis aimed to use fully coupled 3D CFD/FEA models to establish the sensitivity of the transient metal temperatures to accurate modelling of the labyrinth seal.
The research began by setting out to evaluate the CFD modelling technique for honeycomb labyrinth seals to enable modelling of the labyrinth seal under transient conditions and implementation into a state-of-the-art 3D CFD/2D FEA coupled model. Using a test case from Karlsruhe UTC, a 3D CFD model was created based on a benchmarking analysis and sensitivity tests for turbulence model and grid size. The CFD model produced total heat transfer rates within the empirical uncertainty despite limited local heat transfer agreement due to a lack of thermocouple resolution. Important findings on the corner radius and honeycomb wall heat transfer were formed that are critical to modelling the honeycomb labyrinth seal accurately. A recommendation for CFD best practice is provided for grid generation, turbulence modelling, solution setup and convergence practices.
In the next part of the research, the labyrinth seal transient response subject to an initial step change in inlet temperature was investigated. The investigation found that despite significant high frequency, bounded fluctuations in the transient response, transient CFD is unnecessary for engine cycle modelling due to the large fluid to solid time ratio. The values of interest (i.e., the mass flow rate or surface heat transfer) can be represented by mean or pseudo-steady values without incurring the computational cost of transient CFD. Crucial to the understanding of honeycomb stator deterioration, the initial transient heat transfer response of the honeycomb walls shows the potential for producing increased thermal gradients in the honeycomb walls. A 2D conjugate honeycomb labyrinth seal model produced mixed results when emulating the 3D transient response but with adequate calibration, provides a computationally inexpensive method for retrieving values of interest.
In the third part of the research, an engine labyrinth seal was modelled over the onset of slam deceleration in both steady and unsteady CFD and compared to the Rolls Royce 1D analysis tools. The comparison showed that the 1D correlation method of the Rolls Royce analysis tools overpredicted the labyrinth seal fluid temperature rise as well as producing a non-physical heat transfer distribution on the labyrinth seal surfaces. The unsteady CFD did not show any significant transient phenomena under realistic transient conditions over a short time frame of the order of the deceleration time.
In the final part of the research, a coupled 3D CFD/2D FE model is produced to investigate the impact of honeycomb labyrinth seal modelling on the
aero-thermomechanical analysis. For simplicity, a manual coupling approach has been adopted, and this method converged within four coupling runs. This simpler approach, compared to more complicated automated methods, should be of use to engine designers. Over the first ramp of slam deceleration, the labyrinth seal showed a cooler, more homogenous temperature distribution compared to the standalone FEA model, with a temperature difference that would be important when considering seal design safety and component life cycles. Away from the labyrinth seal, there is limited sensitivity at thermocouple locations and the modelling approach should be developed to provide more understanding of the CFD impact on transient metal temperatures.