Numerical Methods. Seminars.

Universitat Pompeu Fabra

2023-2024

Seminar I. DE and NM basic concepts, Euler’s method and errors.

Problem 1

Heat approximately 100 ml of water in a beaker. Set the beaker on the kitchen table, insulated from the
table surface, where it will not be disturbed. Imagine that the temperature sensor placed in the kitchen
measures 20ºC. Initially the temperature of water is 75ºC.Three minutes later it cools up to 55ºC.

a) Find the expression of the temperature as a function of time, T (t), by solving the appropriate
differential equation.

b) Determine the value of the coefficient of heat transfer, c, and specify its units.

c) Determine which will be the temperature a quarter of an hour later.

Problem 2

Let y ′ = cos x, y(0) = 1 be the initial value problem.

a) Compute the exact value of y(0.2) with four decimals.

b) Compute the approximated value of y(0.2) using the Euler method with a time step h = 0.05. Write
all the intermediate steps. Round to the fourth decimal place.

c) Compute the approximated value of y(0.2) using the Euler method with a time step h = 0.025. Write
the three first intermediate steps. Round to the fourth decimal place

d) Compute the relative error of y(0.2) obtained in (b) and (c). Critically evaluate the need or not to
reduce the step h in this case.

1
 Numerical Methods. Seminars.

Universitat Pompeu Fabra

2023-2024

Seminar II. Runge-Kutta method.

Remember that knowledge is cumulative and therefore at any time you may be asked to apply concepts
studied in previous topics.

Problem 1

Let f (t, x) = 2tx, x(1) = 1 be the initial problem value.

REMARK 1: Truncate at the third decimal for all sections.

REMARK 2: Explain the procedure.

a) Compute the expression of x(t) and the exact value of x(1.5).

b) Compute the approximated value of x(1.5) using the Euler’s method with a time step h = 0.1. Write
all the intermediate steps.

c) Compute the approximated value of x(1.5) using the Runge-Kutta 2 method with a time step h = 0.1.
Write the formula for xk+1 and the first three intermediate steps but compute all of them.

d) Compute the approximated value of x(1.5) using the Runge-Kutta 4 method with a time step h = 0.1.
Write the first three intermediate steps but compute all of them.

e) Are the approximated values of x(1.5) equal? Describe what is happening. Explain and justify your
answer. How can you improve the worst approximation?

2
 Numerical Methods. Seminars.

Universitat Pompeu Fabra

2023-2024

Seminar III. Explicit Methods for 1D+t PDEs.

Iterative Numerical Solution of Linear Systems.

Problem 1

Consider the system of equations:

7x1 − 2x2 + 3x3 = 19

x1 + 6x2 − 2x3 = 8
−x1 − 2x2 + 7x3 = 15

This system of equations has as unique solution x̄ = (2, 2, 3)t .

(k+1) (k+1) (k+1)

a) Write the general formula of the Jacobi’s method for x1 , x2 and x3

b) Using x¯0 = (0, 0, 0)t as initial guess, calculate the first three iterations in detail with the Jacobi’s
method. Remark: round to the 3rd decimal.

(k+1) (k+1) (k+1)

c) Write the general formula of the Gauss-Seidel’s method for x1 , x2 and x3

d) Using x¯0 = (0, 0, 0)t as initial guess, calculate the first three iterations in detail with the Gauss-Seidel’s
method. Remark: round to the 3rd decimal.

e) Given the results obtained in the 3rd iteration with both methods, check if the stopping criteria
||x̄(k+1) −x̄(k) ||
||x̄(k+1) ||
< 10−1 is reached with any of them. Which method has faster convergence? Justify
your answers.

3
Problem 2

A string of length L = 120 cm attached at both extremities is held at its median point and then released
m
at time t = 0. Its dynamics is modelled with the 1D wave equation with coefficient c = 10 s . The height
of the rope is monitored every 20 ms at 5 equally spaced points between 0 and 120 cm. At the initial
instant, the height at these points is, respectively, 0.0, 0.5, 1.0, 0.5 and 0.0 (as shown in the table below).
Remark: for all the calculations, round to the 3rd decimal

a) Draw the system considering all the given initial conditions.

b) Which are the boundary conditions? Has the system any symmetry?

c) Compute the value of µ

d) Compute the height of the string at the 5 different points across the first 140 ms using finite differences
method (explicit formulation) with time step 20 ms. Fill the following table with your results.

tj -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

(j)
h1 0.000 0.000
(j)
h2 0.500 0.500
(j)
h3 1.000 1.000
(j)
h4 0.500 0.500
(j)
h5 0.000 0.000

(j)
Where hi = h(xi , tj ) is the height of the rope at the position xi at time tj (that is, for the iteration
j).

e) Describe the movement of the rope over time. How long does it take to reach maximum negative
amplitude ? Estimate the frequency of the system and compare with the theoretical value.

4
 Numerical Methods. Seminars.

Universitat Pompeu Fabra

2023-2024

Seminar IV. Implicit Methods for 1D+t PDEs.

Problem 1

We have a 4 cm rope fixed at its ends. We want to know the height of the rope (h) at 5 different positions
(note that two positions are the fixed ends). At time 0, we stretch the string through the centre 1 cm (we
consider that the other positions are at 0). The initial velocity is 0 m/s, but according to the properties
of the string, the wave will be propagated with velocity c = 0.1m/s. Write the systems of equations that
will describe the height of the string at those 5 positions at time 0.2 with a step of 0.1 using the implicit
finite difference method.

Problem 2

An ”infinity pool” or ”vanishing edge pool” is a type of swimming pool design where the edges appear to
merge with the surrounding landscape, giving the illusion that the water extends infinitely into the horizon.
The water in an infinity pool flows over the edge into a hidden trough, which collects and recirculates the
water back into the pool. The surface of a very thin line of water between two edges can be modelled as
a wave (see Figure 1): At time zero (m = 0), the height of the wave at 6 positions xj separated by 0.1m
is 0.00, −1.00, 1.00, −1.00, 1.00, 0.00, (consider j = 0 and j = 5 to be at the edges of the pool). The initial
√
velocity is one. According to the properties of waves, c = 2 m/s. Write the system of equations of the
Crank-Nicolson method at the second iteration (m = 2), when t = 0.2 and with time step 0.1.

Figure 1: 1D scheme of the water wave at t = 0

5
 Numerical Methods. Seminars.

Universitat Pompeu Fabra

2023-2024

Seminar V. 2D+t PDEs.

Problem 1

We have an extremely thin heating pad with a square shape, and the length of its side is L = 16 cm. At
the start of the experiment (t0 = 0s), the pad is at T = 25K. The pad is heated by a resistance located at
the perimeter of the squared pad, and by diffusion, the heat is transferred through the heating pad. The
resistance remains at T = 100K during the experiment.

• QUESTION 1: Discretise the space into a 5 x 5 numerical mesh. Draw the mesh. Indicate the
position of the resistance. (1 pt)

• QUESTION 2: Which are the boundary conditions? Why? Adapt the explanation to the problem.
(1 pt)

• QUESTION 3: Using the explicit finite difference method calculate the temperature at time t1 =
0.2 s at one of the two diagonals of the mesh points (the diagonal between the mesh points 0,0 and
5,5). Consider a conductivity constant of c = 0.008 WK−1 m−1 . (3 pt)

Problem 2

We have an extremely thin heating pad that has a square shape, and the length of its side is L = 10 cm.
At the start of the experiment (t0 = 0s), the pad is at T = 25K. The pad is heated by a resistance located
at the centre of the pad, and by diffusion, the heat is transferred through the heating pad. The resistance
remains at T = 100K during the experiment. The edges of the pad are covered by a foam material that
provides thermal insulation from the surrounding environment.

• QUESTION 1: Discretize the space into a 5 x 5 numerical mesh. Indicate the position of the
resistance. Draw the mesh. (1 pt)

• QUESTION 2: Which are the boundary conditions? Why? Adapt the explanation to the problem.
(1 pt)

6
• QUESTION 3: Write the system of equations for the implicit finite difference method at time
t = 0.1 s at the inner points (mesh points not at the boundary). Consider a conductivity constant
of c = 0.15 WK−1 m−1 . (3 pt)