Skip to main navigationSkip to main content
The University of Southampton

Research project: Mixing cheap and expensive simulations in an optimisation

Currently Active: 

Traditionally automated design optimisation methods use the results from computational simulations to drive an improvement in a design. However, the expense of such simulations can considerably restrict the scope of an optimisation. An alternative is to mix expensive and cheap simulations within a single design optimisation to improve performance.

Coarse and fine meshes
Compressor rotor

The best simulation is obviously the one which includes the most detail, right? To some extents this is true, but such simulations are generally very expensive to perform even with today's computing power. Using such simulations within a design optimisation, whilst possible, can result in each optimisation taking a considerable amount of time and only a small number of parameters can be varied.

One way around this is to combine expensive simulations with cheaper simulations. These cheap simulations whilst not as accurate as their expensive cousins help to generally point an optimiser in the right direction and can considerably improve optimisation performance for very little additional computational effort.

Temperatures from a simulation of a whole gas turbine engine
Gas turbine engine temperature

The Computational Engineering & Design Group has been involved in the development of methods to mix cheap and expensive simulations within an optimisation loop. To date these methods have been successfully applied to:

  • Mixing converged and partially converged aerodynamic simulations
  • Mixing aerodynamic simulations of varying mesh density and geometric complexity
  • Mixing aerodynamic simulations with different boundary conditions
  • Mixing steady-state and transient structural simulations of whole engines

Related research groups

Computational Engineering and Design

Key Publications

Share this research project Share this on Facebook Share this on Twitter Share this on Weibo
Privacy Settings