Advances of turbomachinery design optimization

Automatic design methods are changing the approach to and processes involved in gas turbine design. These methods can be used on a global scale to explore design space, on a local scale as tools in the specialized design of engine components, or as supporting methods in the optimization of existing design methods. For these purposes, optimization methods may be deployed as the primary design tool or hybridized with other automatic design methods to find new ways to explore the design space. In this paper, three novel examples are presented to demonstrate each of these ways of designing via automation. In the first study, a state-of-the-art inverse design method is used to design a compressor stage. The calculation achieves multiple design targets of streamwise loading distribution, stage pressure ratio, and stage exit ow angle, all radially varying, in addition to massflow. These targets can be distributed across two operating points without compromising the ability to satisfy them. For the first time, a genetic algorithm is wrapped around this inverse design code to form a hybrid automatic design method which is used to optimize for improved aerodynamic efficiency and stall margin, and demonstrate the potential for useful hybridization of automatic techniques in turbomachinery design. In the second study, a very large-scale eddy simulation is used to simulate the ows around the cutback trailing edges of high-pressure turbine blades. For a given external blade design and mainstream ow, a genetic algorithm was used to control the progress of the optimization, aimed at improving the layout of the internal structures within the blade. The genetic algorithm was run for ten generations, by which time, the parameter of fitness-an idealised measure of film cooling performance-was found to have improved significantly over the initial precursor generation. The third study shows the adjoint based improvement of multi-block structured meshes for CFD in several engine parts. The form of the block structure used for complex domains considerably affects the quality of the mesh, which necessarily has a significant knock-on effect on the quality of CFD design. It is normally unclear which blocking would yield the optimal mesh for a specific geometry. Here, the adjoint methods typically employed in design optimization can be used to decide on the mesh block structure. As well as showing the above examples, the paper finally explores some future aspects of design optimization and in particular how eddy-resolving simulations might be used in design optimization in say the next 10 years.