Citation
Mohd Ijam, Hazizah
(2020)
Adaptive step size of diagonally implicit block backward differentiation formulas for solving first and second order stiff ordinary differential equations with applications.
Doctoral thesis, Universiti Putra Malaysia.
Abstract
In this thesis, new classes of block methods based on backward differentiation formula (BDF) for
solving first and second order stiff ordinary differential equations (ODEs) are developed. These
methods are implemented in diagonally implicit
structure and generated the solutions of yn+1 and yn+2 simultaneously in a block. The formulas are
constructed by taking a non zero arbitrary, incorporating a free
parameter, ρ and hence producing ρ diagonally implicit block backward differ- entiation formula (ρ
DIBBDF) which contain the block backward differentiation formula (BBDF) as a subclass.
Initially, the derivation of ρ DIBBDF in fixed and adaptive step approaches for the solution of
first order stiff ODEs have been described. The classes have the advantage of producing a different
set of formulas that possess A-stability properties by selecting the ρ value within the interval
(-1,1). The order, consistency, zero stability, absolute stability and stability region for the
methods have been determined to ensure their applicability in solving the stiff ODEs. The numerical
results have marginally better performance for the fixed step formula and competitive achieve- ment
for the adaptive step formula when compare to the existing BBDF methods.
To deal with the system of second order stiff ODEs, ρ DIBBDF is formulated suited well with the
systems in its original form, without transforms it to the first order ODEs. The convergence and
stability properties also have been analyzed. ls for the method have been obtained and their
stability regions have been discussed. The methods are implemented in fixed and adaptive step approaches.
Comparisons on numerical results to existing BBDF methods demonstrate a comparable performance of
both fixed and adaptive step formulas in terms of accuracy.
The ρ DIBBDF algorithms are written in C programming language. All the ap- proximate solutions of
the standard problems and application systems of stiff ODEs generated by ρ DIBBDF agrees well with
the exact solutions and approximate solutions computed by Matlab stiff solvers. All developed
methods with ρ = 0.75 have shown to perform the computational work in a lesser time when compared
to the existing BBDF methods of the corresponding order.
In conclusion, the proposed methods have shown the suitability and reliability to solve linear and
non-linear systems in different level of stiffness with comparison to the existing BBDF methods and
Matlab stiff solvers. Thus, the new methods developed can be included as viable alternatives for
solving first and second order
stiff ODEs.
Download File
Additional Metadata
Actions (login required)
|
View Item |