Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.694449
Title: Meta-modelling of intensive computational models
Author: Bashar, Hasanain
ISNI:       0000 0004 5991 5280
Awarding Body: University of Sheffield
Current Institution: University of Sheffield
Date of Award: 2016
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
Engineering process design for applications that use computationally intensive nonlinear dynamical systems can be expensive in time and resources. The presented work reviews the concept of a meta-model as a way to improve the efficiency of this process. The proposed meta-model will have a computational advantage in implementation over the computationally intensive model therefore reducing the time and resources required to design an engineering process. This work proposes to meta-model a computationally intensive nonlinear dynamical system using reduced-order linear parameter varying system modelling approach with local linear models in velocity based linearization form. The parameters of the linear time-varying meta-model are blended using Gaussian Processes regression models. The meta-model structure is transparent and relates directly to the dynamics of the computationally intensive model while the velocity-based local linear models faithfully reproduce the original system dynamics anywhere in the operating space of the system. The non-parametric blending of the meta-model local linear models by Gaussian Processes regression models is ideal to deal with data sparsity and will provide uncertainty information about the meta-model predictions. The proposed meta-model structure has been applied to second-order nonlinear dynamical systems, a small sized nonlinear transmission line model, medium sized fluid dynamics problem and the computationally intensive nonlinear transmission line model of order 5000.
Supervisor: Harrison, Robert F. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.694449  DOI: Not available
Share: