COMPUTATIONAL FLUID DYNAMICS: ASSIGNMENT PROBLEMS
Dr K M Singh, Indian Institute of Technology Roorkee NPTEL VI.1
MODULE VI
SOLUTION OF ALGEBRAIC SYSTEMS: ASSIGNMENT
VI.1. TDMA can be used for solution of the algebraic system of equations arising from the finite
difference discretization of
(a) one dimensional Laplace or Poissons equation
(b) two dimensional Laplace or Poissons equation
(c) three dimensional Laplace or Poissons equation
(d) multi-dimensional heat conduction equation if ADI method is used.
VI.2. Consider the sparse linear algebraic system Ax = b obtained from central difference
discretization of three dimensional Poisson equation. Which of the following would be the
most suitable solution method(s) when the number of grid points is very large?
(a) one dimensional Laplace or Poissons equation
(b) two dimensional Laplace or Poissons equation
(c) three dimensional Laplace or Poissons equation
(d) multi-dimensional heat conduction equation if ADI method is used.
VI.3. For a sparse linear algebraic system Ax = b , which iterative solver amongst the following
would exhibit the fastest convergence?
(a) Jacobis method.
(b) Gauss-Seidel method.
(c) SOR method.
(d) GMRES method.
VI.4. Which of the following iterative methods can be used for the solution of asymmetric linear
systems?
(a) Bi-conjugate gradient method.
(b) Conjugate gradient method.
(c) SOR method.
(d) GMRES method.
VI.5. The most effective method for the solution of the sparse linear algebraic system Ax = b
obtained from central difference discretization of one dimensional Poisson equation is
(a) LU decomposition.
(b) Conjugate gradient method.
(c) Multigrid method.
(d) TDMA.
VI.6. The convergence of the preconditioned conjugate gradient iterations depends on the
preconditioner. Which of the following statements is true with respect to convergence rate of
PCG iterations while comparing Jacobis (diagonal) preconditioner and incomplete Cholesky
preconditioner?
(a) Diagonal preconditioner is better than incomplete Cholesky preconditioner.
(b) Incomplete Cholesky preconditioner is better than the diagonal preconditioner.
(c) Both provide the same rate of convergence.
(d) None of the above.
COMPUTATIONAL FLUID DYNAMICS: ASSIGNMENT PROBLEMS
Dr K M Singh, Indian Institute of Technology Roorkee NPTEL VI.2
VI.7. Use of finite difference or finite volume method for the solution of generic transport equation
for transport of a scalar yield a sparse linear system of equation given by
A= Q
where A is the system matrix, is the vector of nodal unknowns, and Q is the known load
vector. Iterative solvers are usually preferred for solution this sparse system. Provide the
basic equations (algorithms) for the following basic solvers
(i) Jacobi Method
(ii) Gauss-Seidel Method
(iii) SOR method
Which of these three would exhibit the best convergence behavior? Would you use any of
these for solution of very large (say, order of a million) system?
VI.8. Provide the definition of projection method for the solution of a linear system. What is a
Krylov subspace method? What is general form of the approximate solution obtained from a
Krylov subspace method? For which choice of the constraint subspace does this method
reduce to conjugate gradient method, and GMRES. Write the algorithm for the
preconditioned conjugate gradient (PCG) method. What are the conditions on the system
matrix A for application of conjugate gradient method? What are the situations in which
GMRES (or Bi-CG) must be used?
VI.9. Consider the finite difference (or finite volume) solution of 2-D steady state heat conduction
problem governed by the PDE
2 2
2 2
0
T T
x y
+ =
using central difference scheme on a uniform Cartesian grid. Outline an iterative solution
procedure for solving the resulting penta-diagonal system of equations using TDMA (tri-
diagonal matrix algorithm). Note that such an algorithm is popularly referred as ADI or
alternating sweep-direction method.
VI.10. Can you extend the ADI type method of the previous problem for finite difference solution of
3-D steady state heat conduction problem
2 2 2
2 2 2
0
T T T
x y z
+ + =
