Geometry parameterisation and response surface-based shape optimisation of aero-engine compressors

Sammanfattning: The main goal in jet engine design is to achieve light, compact and highly efficient systems while reducing the cost and design times. Traditionally, the design has been based on trial-and-error procedures, which count on the skills and experience of the designer to suggest design improvements. In step with recent advances in efficient and robust computational algorithms and the fast development in computer resources, computational fluid dynamics (CFD) has matured to a stage where it provides substantial insight into the physical processes involved in engineering flows. The CFD solvers do, however, not indicate which geometry modifications are required to improve performance. Even if it is possible to investigate a very large number of design alternatives it is difficult to find the optimum only by trial-and- error and it can also be very time consuming. The search for an improved design can be aided by the integration of a numerical optimisation algorithm with the CFD code. Before this can be done several decisions have to be taken. The first is how to make a parameterisation of the geometry to describe allowable shape variations in the design. The second decision is to define problem constraints. Finally, the performance parameters to be used in the definition of the objective function have to be identified. The main objective of this work has been to develop and implement methods for shape optimisation of compressor blades. A large part of the work has been devoted to the development of a compressor blade design system. The geometry description is based on Non-Uniform Rational B-Splines (NURBS), which are used by modern computer aided design tools and have suitable mathematical properties for parameterisation of fluid flow domains. The problem with the NURBS parameters is that they do not coincide with the traditional design parameters used by the aerospace engineers. However, keeping the traditional parameters as design variables will help the designer to chose fewer but more relevant design parameters for the problem. A separate optimisation algorithm has thus been integrated in the blade design system to optimise the NURBS parameters to achieve the desired traditional design parameters, also called implicit design parameters. A second method to reduce the total number of design parameters is to introduce B-spline mapping curves that interpolates shape modifications of a starting blade in the spanwise (radial) direction. This method allows the designer to arbitrary select the number of degrees of freedom depending on his/her needs. The blade design system has been used to optimise a compressor blade geometry via construction of response surface approximations. Optimisation both with and without implicit design variables has shown that using implicit design variables can substantially reduce the number of CFD simulations. The response surface method has shown to be efficient in filtering numerical noise intrinsic to the numerical data produced by the CFD code. The method is also capable of handling multi-criterion optimisation problems in a straightforward manner by building a composite response surface from individual response surfaces.