Fundamentals of Numerical Thermo-Fluid Dynamics 322.
061
Examples for home preparation
Exercise 2: Finite Difference Method
To be presented on May 13, 2020
2.1 Determine the coefficients a to d of the central finite difference scheme for a uniformly
spaced grid (∆x = const.):
∂u
= auj−2 + buj−1 + cuj+1 + duj+2 , (1)
∂x j
using Taylor series expansion. What is the order of accuracy of the scheme? Is the
scheme diffusive or dispersive?
Notation: uj ≡ u(xj ) ≡ u(j∆x)
2.2 Determine the coefficients a and b of the following central finite difference scheme for
a NON-UNIFORM grid (∆xj = xj+1 − xj 6= const.):
∂u
= auj−1 + buj+1 . (2)
∂x j
What is the order of accuracy of this scheme? Would the order of accuracy change
for a uniform spacing? Can you conclude any recommendations for designing a non-
uniform finite-difference grids?
Notation: xj−1 = xj − ∆xj−1 , xj+1 = xj + ∆xj
2.3 For the function u = sin(πx) compute the first derivative du/dx at x = 0.4 using the
schemes below:
du uj+1 − uj
(a) ≈ ,
dx ∆x
du uj+1 − uj−1
(b) ≈ ,
dx 2∆x
du uj−2 − 8uj−1 + 8uj+1 − uj+2
(c) ≈ ,
dx 12∆x
and compare the results with the exact solution. For each scheme, plot the depen-
dence of the discretization error on ∆x in logarithmic scale. Can you determine the
order of accuracy of the scheme from the plot?
1
Exercise 2 322.061: Fundamentals of Numerical Thermo-Fluid Dynamics
2.4 Consider the solved example Solution of 1D steady diffusion. Reduce the number
of grid points N to check how many grid points you need to approximate the exact
solution with a maximum relative error of 1%. Then change the source term to f = 1
and repeat. Explain your observation.
2.5 Consider the Friedrich’s equation
d2 f df
2
+ = a on Ω = [0, 1] (3)
dx dx
f = 0 at x = 0 (4)
f = 1 at x = 1 (5)
(a) Assume that f is a passive scalar (e.g. temperature or chemical concentration)
transported in a fluid. Explain the physical meaning of each term of (3).
(b) Discretize (3) using finite differences for uniformly spaced grid.
(c) Solve the problem numerically for = 0.1 and a = 2.
(d) What would be the main problem if gets too small? How would you approach
such case?
