Robust Aerodynamic Design Optimisation of Compression Systems

Research Theme: Computational Design

The design of a new compress ion system for aircraft application is a complex task: design objectives are almost always conflicting, the design space is large, nonlinear and highly constrained. It is almost impossible, even for an experienced designer, to obtain an optimal design by simple trial and error: more sophisticated tools for exploring the design space are needed to obtain a complete overview.

Motivation

Improve the aerodynamic design of compression systems.

Objectives

  • Closer integration of preliminary and detailed design
  • Simultaneous optimisation of different engine components
  • Minimisation of computational time required
  • Design of robust components

Method

Meta-heuristic optimisation tools can improve the preliminary design of turbomachinery, allowing a more thorough and rapid exploration of the design space to identify its most promising regions that will then be verified and further analysed with higher fidelity tools.

Findings

Figure 1 shows the result of the two-objective (efficiency and surge margin) optimisation of a 7-stage intermediate pressure compressor preliminary design. Starting from an initial geometry (AQ7), two different optimisation strategies were applied to the problem: a modified gradient-based method and the Tabu Search algorithm. The compressor geometry was parameterised through 91 variables and 44 of these were varied during the optimisation process (mean line shape and area distribution, pressure ratio for each stage, stator exit blade angle, cord lengths and number of blades). The performance was analysed with in-house compressor performance prediction software based on mean line correlations.

The results are compared with the original compressor (AQ7) and an improved geometry obtained in a previous research with a mix of human-based design and gradient-based optimisation within a 5-variable design space. The improvements obtained with the gradient-based optimiser demonstrate the significant size of the design space and thus the need for an automated preliminary design process that is able of analysing a large number of designs in the shorter possible time, while the distance between the two Pareto fronts demonstrates the importance of a stochastic approach in a multi-variate, multi-modal and multi-objective optimisation problem such as the preliminary design of compressors.

Additional constraints were also analysed: a maximum allowed Koch factor was set to increase the confidence in the predicted surge margin level and the exit Mach number was limited to the initial design value to avoid excessive loadings in the following engine components (figure 2).

Acknowledgements

Support for this project was provided by RAS, the Cambridge European Trust and the EPSRC.