The ultimate Gas-Turbine-Life-Model would be one that could predict the life of every component in the engine, based on current engine performance. Thus, during a flight, the mechanical life of all the components could be monitored. This implies that all the necessary calculations would have to be performed in real-time. There is no such model available at present.
The models that do exist, are those that obtain engine performance history from in-flight monitored engine parameters and flight conditions. The data is then downloaded for processing after the completion of a flight. ARMSCOR, along with universities in South Africa, is currently developing a gas turbine life model. The goal of the research presented in this dissertation, was to demonstrate a modelling procedure for the calculation of realistic transient thermal stresses. This, in turn, provides quality input data for thermal fatigue calculations, for the most critical component in the engine, namely the NGV.
Computing the thermal stresses inside a turbine blade is no small task. Simulating turbine blade thermal stresses requires both flow and structural analysis for the blade. The flow and structural analysis have to be performed separately, due to each employing different formulations for the governing equations. FLUENT, which uses the Finite Volume formulation, was used to resolve the flow field while NASTRAN, which uses Finite Element formulation, was used for the structural analysis.
The understanding of thermal stress behaviour begins with performing a steady state analysis. The CFD model must be validated against experimental data, while choosing a suitable code for the FEM model, is vital. This is due to the limitations of performing experimental stress measurements. Only once the steady state model has been validated can transient simulations be performed. For the transient models, the choice of boundary conditions is important. The unsteady TIT, being the root cause of thermal stresses, must be modelled correctly in order to accurately model the corresponding unsteady thermal stresses.
The transient simulations were performed for cases with abrupt temperature changes, which result in thermal shocks, and cases that represent typical operational flight data, which can be used for thermal fatigue and crack growth calculations.
The results of the steady state aerodynamic analysis were in excellent agreement with the experimental data of Nealy et. al. (1984) and Bohn et. al. (1995). The flow field was accurately resolved, which resulted in the surface pressure prediction being exact with that of Nealy's experimental data. All three turbulence models investigated gave the same excellent prediction. Grid independent solutions were easily achieved with all the mesh models producing the same results.
The results of the surface temperature distribution were far more difficult to achieve. Grid independent solutions were only achieved when a total decomposition of the flow field was performed, along with continuous mesh refining in regions where high pressure and temperature gradients were found. All three turbulence models performed exceptionally well in the turbulent regions of the blade, where the Spalart-Allmaras model over predicted Bohn's experimental data by roughly 1 %. All the turbulence models showed an over prediction in excess of 10 % in the Laminar region. It was thus shown that none of the FLUENT turbulence models could accurately predict the heat load to the blade, where this fmding is strongly emphasised throughout the open literature.
After the extensive effort of predicting the temperature distribution using the most established turbulence models, a Laminar model was used for the analysis. It was discovered, that the Laminar model could highly accurately predict the surface temperature distribution. The model over predicted Bohn's experimental data by a mere 1 % in the laminar regions of the blade. The results of the Spalart-Allmaras and Laminar model were combined, which resulted in the validation of the CFD model.
The results of the steady state FEM analysis were compared to the predicted results of Bohn.
The thermal analysis resulted in the internal blade temperature profile, which over predicted Bohn's data by 2 % - 3 %. The mechanical analysis predicted the correct thermal stress contour trends, but under predicted the stress magnitudes by an average of 17 % for the entire blade. The reason for the under prediction is believed to be due to different solvers used for the two predictions, as explained in chapter 5. For the FEM analysis, the mesh used is not nearly as important as it is for the CFD analysis. The correct material and element characterisation and accurate results form the aerodynamic analysis are most vital to the accuracy of the stress prediction.
Transient stress profiles were calculated for the T56 engine, where the stress curve is primarily dependent on the changes in TIT. From the stress plots it can be argued that based on the stress behaviour as explained in chapter 6, the model is accurate and represents a
realistic analysis. The thermal shock during start up was modelled and showed that the resulting stress rose tremendously, which causes the most damage to the component. The minimum and maximum stresses are presented and can be used as boundary conditions to calculate the critical number of cycles to failure.
The success of the transient thermal stress plots is mainly due to the discovery of the Laminar model. The accuracy of the aerodynamic analysis is the heart of the work presented in this dissertation. The combination of the Spalart-Allmaras and Laminar model for the aerodynamic analysis proves to be an accurate tool, for predicting the heat load as well as the corresponding thermal stresses. This combination thus proves to be a solution to the temperature prediction dilemmas discussed in the literature review.