Robust Aerodynamic Design Optimisation of Compression Systems

Year: 2009
Author: Ghisu, Tiziano
Supervisor: Parks, Geoff
Institution: University of Cambridge
Page(s): 274


Designing a new gas turbine is a di cult task, not only due to the intrinsic complexity of the phenomena involved, but also to the large number of design variables which, coupled with a number of conflicting objectives and constraints, generate a highly fragmented and multi-modal design space, making it impossible, even for an experienced designer, to locate the optimal solution by simple trial and error. This complexity has led to a hierarchical, modular and iterative design process, where not only are the different modules designed separately and at a gradually increasing level of fidelity, but also the di erent disciplines are dealt with by separate teams of specialists. This approach gives a fundamental importance to choices made very early during the design process, when little knowledge about the detailed geometry exists, encouraging the use of conservative solutions, with the risk of concealing important trade-o s between modules and disciplines and favouring the convergence to sub-optimal designs. As a first step towards reducing the level of decomposition typical of gas turbine engines' design, this work concentrates on producing an integrated aerodynamic preliminary design system for a core compression system, thereby reducing the number of “artificial constraints” required from the conceptual design phase. These interfaces, and possibly more design variables, will necessarily be xed before the detailed design, but, having postponed the setting of some global design variables to a point when more information is available, with a lower impact on the final result. The size and complexity of the design space, together with the ever present need for shorter design times, makes the use of an intelligent search algorithm essential: Tabu Search was selected for this study as it proved particularly efficient in a number of similar problems. In recognition of the large design space and of the signifficant time required by an algorithm based on a pattern search for local exploration, an improvement based on the idea of Principal Components' Analysis, from the field of Pattern Recognition, was implemented and proved essential for a more rapid and thorough exploration of the design space. While the direct use of CFD in the aerodynamic design of an entire module or, even worse, the whole engine is still infeasible, an intelligent use of higher-fidelity tools can be essential to eliminate the need for conservative solutions required by lack of knowledge or reliable design rules: an example is given in this study, where a two-dimensional in-house flow solver was used within the preliminary design process to remove the need for a conservative inter-compressor duct and to integrate its design with that of intermediate and high pressure compressors. Surrogate models were employed to reduce evaluation times and allow the use of this more expensive tool in a preliminary design time frame. In recognition of the importance of off-design conditions and of the tendency of heavily optimised products to display severe off-design performance degradation, two ways of embedding off-design performance evaluations within the optimisation process (interval analysis and Polynomial Chaos) were suggested and implemented, leading to solutions displaying improved behaviour over a wide range of operating conditions (robust solutions) and further contributing to reduce the level of fragmentation typical of compression systems design. The integrated design system developed in this work demonstrates the potential for improving the preliminary design process for compression systems, both in terms of design-point and off-design performance, while also minimising process duration through the use of an automatic design approach and of intelligent search algorithms.

Please sign in to your account

This site uses cookies and other tracking technologies to assist with navigation and your ability to provide feedback, analyse your use of our products and services, assist with our promotional and marketing efforts, and provide content from third parties. Privacy Policy.