Use this URL to cite or link to this record in EThOS:
Title: Coefficient identification problems in heat transfer
Author: Hussein, Mohammed Sabah
ISNI:       0000 0004 5923 5330
Awarding Body: University of Leeds
Current Institution: University of Leeds
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Access from Institution:
The aim of this thesis is to find the numerical solution for various coefficient identification problems in heat transfer and extend the possibility of simultaneous determination of several physical properties. In particular, the problems of coefficient identification in a fixed or moving domain for one and multiple unknowns are investigated. These inverse problems are solved subject to various types of overdetermination conditions such as non-local, heat flux, Cauchy data, mass/energy specification, general integral type overdetermination, time-average condition, time-average of heat flux, Stefan condition and heat momentum of the first and second order. The difficulty associated with these problems is that they are ill-posed, as their solutions are unstable to inclusion of random noise in input data, therefore traditional techniques fail to provide accurate and stable solutions. Throughout this thesis, the Crank-Nicolson finite-difference method (FDM) is mainly used as a direct solver except in Chapter 7 where a three-level scheme is employed in order to deal with the nonlinear heat equation. An explicit FDM scheme is also employed in Chapter 10 for the two-dimensional case. The inverse problems investigated are discretised using the FDM and recast as nonlinear least-squares minimization problems with simple bounds on the unknown coefficients. The resulting problem is efficiently solved using the \emph{fmincon} or \emph{lsqnonlin} routines from MATLAB optimization toolbox. The Tikhonov regularization method is included where necessary. The choice of the regularization parameter(s) is thoroughly discussed. The stability of the numerical solution is investigated by introducing Gaussian random noise into the input data. The numerical solutions are compared with their known analytical solution, where available, and with the corresponding direct problem numerical solution where no analytical solution is available.
Supervisor: Lesnic, Daniel Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available