Surrogate Modelling for Power Plant Design, a talk by Dr Sarah Davis (Alstom)
Abstract
In order to maximise market share engineering companies continuously strive to design the best product at the lowest cost. Often there will be tens or even hundreds of design parameters that determine the performance of the end product.
Performing an optimisation algorithm based search – or brute force grid search – of the design space may be useful in finding the optimum design in situations where there are only a few variables. However these approaches would require infeasibly large numbers of evaluations for high dimensional situations. Even for five or ten parameters, it is not realistic to use optimisation algorithms due to the computational expense of one run, of, for example, a computational fluid dynamics simulation, which may take minutes or hours to complete.
A solution to these problems is to use a limited number of sample points to generate a surrogate model of the original system, the behaviour of which is expensive to determine.
In this talk I will introduce some surrogate modelling techniques and discuss some of the issues faced by industry when trying to apply them to the product design process.
Dr Sarah Davis’ Thesis was “Crepant resolutions and A-Hilbert schemes in dimension four” with Professor Miles Reid at the University of Warwick.
No charge is made to attend meetings and non-members are welcome.
