Design Space Exploration

Research Theme: Process Management

A compressor blade is a complex geometry and the search for better solution is a very challenging task. The aims of this project are:

  • to generate regression models able to capture the complexity of the design space,
  • to give the designer a tool able to drive the search of feasible solutions in new areas of the design space.

Motivation

The need to speed up the process of designing more efficient compressor blades, by maximising a number of objective functions.

Objectives

  • To improve the preliminary analysis of blade design
  • To build a reduced order model for the blade, in such a way that the useof expensive CFD tools is limited
  • To build a more efficient methodology to explore the design space.

Method

In the first part of the project, design of experimental techniques have been reviewed and their pros and cons analysed. After sampling a relatively large number of blade designs in the design space, Proper Orthogonal Decomposition was used to project the data on a 3D space, thus reducing the complexity of the initial model.

This graphic representation (figure 1) also makes it possible to visualise the explored multidimensional design space. The model can be used as a surrogate of the original one, and response surfaces on selected areas of the design space can be easily built.

Additionally, the design charts generated can be used to drive the search of feasible solutions in new areas of the design space.

Findings

The charts show that feasible designs often occur in clusters. Such areas of the design space represent the targets for further exploration.

Acknowledgements