Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.633231
Title: Surrogate model construction for steady aerodynamic loads
Author: Mackman, Thomas James
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2013
Availability of Full Text:
Access through EThOS:
Abstract:
An efficient method for predicting steady aerodynamic loads with respect to varying geometric and flow parameters is to use a surrogate model to interpolate or approximate a finite set of Computational Fluid Dynamics (CFD) simulations. Further improvements to the strategy for constructing the surrogate have the potential to provide more accurate predicted values or to reduce the number of simulations required to achieve a model of sufficient quality. This work was originally motivated by the task of providing data for calculating structural loads for civil passenger aircraft, but is directly relevant for closely related applications such as providing aerodynamic data for flight mechanics analysis, and quantification of race-car aerodynamic performance. The objective at the outset was to develop aspects toward an improved surrogate modelling strategy for predicting aerodynamic data that enables a reduction in the overall simulation budget. To this end, the fundamental topics of adaptive sampling, model parameter tuning, and practical implementation for aerodynamic data have been investigated, with the goal of developing novel methods in each of these areas, and analysing their operation. Details of an adaptive sampling method based on a combination of curvature-adaptive and space-filling components are presented, including recovery of expected behaviour for analytic functions, formulation of the space-filling component, simultaneous addition of update points; and how best to optimise the criterion efficiently for multidimensional problems. An advanced strategy for choosing locally varying interpolation parameters is then presented, which works by optimising a single value to scale a prescribed local distribution of parameters, subject to constraints on the properties of the interpolation matrix. Following this, the use of various physics-based responses to drive the sampling algorithm and techniques for mitigating noise are investigated for application to aerodynamic data.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.633231  DOI: Not available
Share: