Task 1: Discretization and Implicit Euler's Method

Discretization of the Advection-Diffusion Equation:

We discretize the advection-diffusion equation using central finite difference scheme for the diffusive
term and implicit Euler's method for time discretization.
The discretized equation for the internal nodes (excluding boundaries) can be written as:

𝑚+1 𝑚 𝐶𝑖,𝑗 −𝐶𝑖−−1,𝑗 𝐶𝑖,𝑗 −𝐶𝑖,,𝑗−1 𝐶𝑖,+1,𝑗 −𝐶𝑖,𝑗 +𝐶𝑖−1,𝑗 𝐶𝑖,,𝑗+1 −𝐶𝑖,𝑗 +𝐶𝑖,𝑗−1
𝐶𝑖,𝑗 = 𝐶𝑖,𝑗 + ∆𝑡 (𝑢 +𝑣 +𝛤( + ))
∆𝑥 ∆𝑦 ∆𝑥 2 ∆𝑦 2

Where
• 𝑚+1
𝐶𝑖,𝑗 is the concentration at node (I,j) and time tn+1
• 𝑚
𝐶𝑖,𝑗 is the concentration at node (I,j) and time tn
• Δx and Δy are the grid spacings in the x and y directions respectively.
• Δt is the time step size.
• u and v are the velocities in the x and y directions respectively.
• Γ is the diffusion coefficient.
Handling Convective Term:
For the Convective term, we use a hybrid scheme:
𝑢∆𝑡 𝑣∆𝑡
• If the Peclet number Pex and Pey (defined as 𝑎𝑛𝑑 respectively) are both less than
∆𝑥 ∆𝑦
2, we use a central difference scheme.
• Otherwise, we use an upwind scheme.
Only perform the first iteration
Given advection diffusive equation
𝜕𝐶 𝜕𝐶 𝜕𝐶 𝜕2𝐶 𝜕2𝐶
+𝑢 +𝑣 = 𝛤 ( 2 + 2)
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦
Parameters
u=1.5; v=0; 𝛤=0.001
Step 1: Compute Convective term
We will compute |pe| and determine
𝑢∆𝑡
|pe| =
∆𝑥

Ly=1m; Lx=50m
 𝐿𝑥 𝐿𝑦
∆𝑥 = ; ∆𝑦 =
𝑁𝑥 −1 𝑁𝑦 −1

Assume that
No. of grid along x =Nx=100
No. of grid along y=Ny=20
Time Step = ∆𝑡 = 0.1

50
∆𝑥 = = 0.5051
100 − 1

1
∆𝑥 = = 0.0526
20 − 1
Now
(1.5)(0.1)
|Pe| = = 0.297
0.5051

Since |Pe| < 2 we will use the central difference scheme for convective term
𝐶3,2 −𝐶1,2 𝐶2,3 −𝐶2,1
Convective term = 𝑢 +𝑣
2∆𝑥 2∆𝑦

MATLAB Code
function [C_final] = solve_advection_diffusion_task1(D, u, v, dx, dy, dt, Nx, Ny, Nt, cw0, cnL)
% Initialize concentration matrix
C = zeros(Nx, Ny, Nt);

% Set initial condition

% Assuming initial concentration is 0 everywhere
C(:,:,1) = zeros(Nx, Ny);

% Time-stepping loop
for n = 1:Nt-1
% Internal nodes update (excluding boundary nodes)
C_temp = C(:,:,n);
for i = 2:Nx-1
for j = 2:Ny-1
% Convective term (hybrid scheme)
if abs(u) < 2 && abs(v) < 2
convective_x = u * (C_temp(i,j) - C_temp(i-1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j-1)) / dy;
else
convective_x = u * (C_temp(i,j) - C_temp(i+1,j)) / dx;
 convective_y = v * (C_temp(i,j) - C_temp(i,j+1)) / dy;
end

% Diffusive term (central finite difference)

diffusive = D * ((C_temp(i+1,j) - 2*C_temp(i,j) + C_temp(i-1,j)) / dx^2 + ...
(C_temp(i,j+1) - 2*C_temp(i,j) + C_temp(i,j-1)) / dy^2);

% Update concentration using implicit Euler's method

C(i,j,n+1) = C_temp(i,j) + dt * (convective_x + convective_y + diffusive);
end
end
end
% Return final concentration
C_final = C(:,:,Nt);
end
% Define parameters
% Define parameters
D = 0.01; % Diffusion coefficient
u = 1.5; % x-velocity
v = 0; % y-velocity
dx = 0.5051; % x-direction mesh size
dy = 0.0526; % y-direction mesh size
dt = 0.1; % time step size
Nx = 100; % Number of grid points in x-direction
Ny = 20; % Number of grid points in y-direction
Nt = 100; % Number of time steps

% Define boundary conditions

cw0 = @(y) 0; % West boundary condition function
cnL = @(x) 0.1 * (x / Nx); % North boundary condition function

% Solve advection-diffusion problem

C_final = solve_advection_diffusion_task1(D, u, v, dx, dy, dt, Nx, Ny, Nt, cw0, cnL);

% Plot concentration distribution at final time

[X, Y] = meshgrid(1:Ny, 1:Nx);
surf(X, Y, C_final);
xlabel('y');
ylabel('x');
zlabel('Concentration');
title('Concentration Distribution at Final Time (Task 1)');
Results of Task 1

Task 2: Implementation of Boundary Conditions

In Task 2, we implement the Neumann boundary conditions for the east and south boundaries.
𝜕𝐶
• For the east boundary (x=L), the Neumann boundary condition 𝜕𝑥 = 0 is applied
implicitly into the governing equations.
𝜕𝐶
• For the south boundary (y=0), the Neumann boundary condition 𝜕𝑦 = 0 is applied
implicitly into the governing equations.
• For the west boundary (x=0), we apply the given boundary condition explicitly.
• For the north boundary (y=H), we apply the given boundary condition explicitly.

Using Central finite difference scheme

𝐶3,2 −𝐶1,2 𝐶2,3 −𝐶2,1
Convective term = 𝑢 +𝑣
𝜕∆𝑥 𝜕∆𝑦

𝐶3,2 −𝐶1,2 𝐶 −𝐶
Convective term = (1.5) ( ) + (0) ( 2,3 2,1 )
2(0.5051) 2(0.0526)
1.5
= (𝐶3,2 − 𝐶1,2 )
1.0102
 𝐶3,2 −2𝐶2,2 +𝐶1,2 𝐶2,3 −2𝐶2,2 +𝐶2,1
Diffusive term = 𝛤 ( + )
∆𝑥 2 ∆𝑦 2

𝐶3,2 −2𝐶2,2 +𝐶1,2 𝐶2,3 −2𝐶2,2 +𝐶2,1

Diffusive term = 0.001 ( + )
(0.5051)2 (0.0526)2

𝐶3,2 −2𝐶2,2 +𝐶1,2 𝐶2,3 −2𝐶2,2 +𝐶2,1

Diffusive term =0.001( + )
0.2552 (0.00277)
1
To Combine the Convective term and Diffusive term to solve the 𝐶2,2
1 0
𝐶2,2 − 𝐶2,2 =∆𝑡 (Convective term+ Diffusive term)

1 0 1.5 𝐶3,2 −2𝐶2,2 +𝐶1,2 𝐶2,3 −2𝐶2,2 +𝐶2,1

𝐶2,2 − 𝐶2,2 =(0.1) ( (𝐶3,2 − 𝐶1,2 ) + 0.001 ( + ))
1.0102 0.2552 (0.00277)

Apply the all boundary condition according Neumann boundary condition

𝐶1,2 = 0; 𝐶2,1 = 0; 𝐶2,3 = 0

1 1.5 𝐶3,2 −2(0)+0 0−2(0)+0

𝐶2,2 − 0=(0.1) ( (𝐶3,2 − 0) + 0.001 ( + ))
1.0102 0.2552 (0.00277)

1 1.5 𝐶3,2
𝐶2,2 =(0.1) ( (𝐶3,2 ) + 0.001 ( ))
1.0102 0.2552

1
𝐶2,2 =(0.1)((1.4848)(𝐶3,2 ) + 0.003915𝐶3,2 )
1
𝐶2,2 =(0.14848)(𝐶3,2 )
This is the complete first iteration with the provided boundary condition.
MATLAB Code
function [C_final] = solve_advection_diffusion_task2(D, u, v, dx, dy, dt, Nx, Ny, Nt, cw0, cnL)
% Initialize concentration matrix
C = zeros(Nx, Ny, Nt);

% Set initial condition

% Assuming initial concentration is 0 everywhere
C(:,:,1) = zeros(Nx, Ny);

% Define Peclet number

Pe_x = u * dt / dx;
Pe_y = v * dt / dy;

% Time-stepping loop
for n = 1:Nt-1
% Internal nodes update (excluding boundary nodes)
C_temp = C(:,:,n);
for i = 2:Nx-1
for j = 2:Ny-1
% Convective term (hybrid scheme)
if abs(Pe_x) < 2 && abs(Pe_y) < 2
convective_x = u * (C_temp(i,j) - C_temp(i-1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j-1)) / dy;
else
convective_x = u * (C_temp(i,j) - C_temp(i+1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j+1)) / dy;
end

% Diffusive term (central finite difference)

diffusive = D * ((C_temp(i+1,j) - 2*C_temp(i,j) + C_temp(i-1,j)) / dx^2 + ...
(C_temp(i,j+1) - 2*C_temp(i,j) + C_temp(i,j-1)) / dy^2);

% Update concentration using implicit Euler's method

C(i,j,n+1) = C_temp(i,j) + dt * (convective_x + convective_y + diffusive);
end
end

% Apply boundary conditions

% West Boundary (x = 0)
C(1,:,n+1) = cw0(1:Ny);

% East Boundary (x = L)
% Neumann boundary condition: dC/dx = 0
C(Nx,:,n+1) = C(Nx-1,:,n+1);

% South Boundary (y = 0)
% Neumann boundary condition: dC/dy = 0 (implicit)
C(:,1,n+1) = C(:,2,n+1);

% North Boundary (y = H)
C(:,Ny,n+1) = cnL(1:Nx);
end
 % Return final concentration
C_final = C(:,:,Nt);
end
% Define parameters
D = 0.01; % Diffusion coefficient
u = 1.5; % x-velocity
v = 0; % y-velocity
dx = 0.5051; % x-direction mesh size
dy = 0.0526; % y-direction mesh size
dt = 0.1; % time step size
Nx = 100; % Number of grid points in x-direction
Ny = 20; % Number of grid points in y-direction
Nt = 100; % Number of time steps

% Define boundary conditions

cw0 = @(y) 0; % West boundary condition function
cnL = @(x) 0.1 * (x / Nx); % North boundary condition function

% Solve advection-diffusion problem with Task 2 implemented

C_final = solve_advection_diffusion_task2(D, u, v, dx, dy, dt, Nx, Ny, Nt, cw0, cnL);

% Plot concentration distribution at final time

[X, Y] = meshgrid(1:Ny, 1:Nx);
surf(X, Y, C_final);
xlabel('y');
ylabel('x');
zlabel('Concentration');
title('Concentration Distribution at Final Time (Task 2)');

Results of TASK 2:
Task3: MATLAB Implementation
function [C_final] = solve_advection_diffusion_task3(dx, dy, dt, u, v, D)
% Define parameters
Lx = 50; % Length of the domain in the x-direction (m)
Ly = 1; % Length of the domain in the y-direction (m)
Nx = round(Lx / dx) + 1; % Number of grid points in x-direction
Ny = round(Ly / dy) + 1; % Number of grid points in y-direction
Nt = 100; % Number of time steps
% Initialize concentration matrix
C = zeros(Nx, Ny, Nt);
% Set initial condition
C(:,:,1) = zeros(Nx, Ny);
% Define boundary conditions
cw0 = @(y) zeros(size(y)); % West boundary condition function
cnL = @(x) 0.1 * (x / Lx); % North boundary condition function

% Define Peclet number

Pe_x = u * dt / dx;
Pe_y = v * dt / dy;
% Time-stepping loop
for n = 1:Nt-1
% Internal nodes update (excluding boundary nodes)
C_temp = C(:,:,n);
for i = 2:Nx-1
for j = 2:Ny-1
% Convective term (hybrid scheme)
if abs(Pe_x) < 2 && abs(Pe_y) < 2
convective_x = u * (C_temp(i,j) - C_temp(i-1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j-1)) / dy;
else
convective_x = u * (C_temp(i,j) - C_temp(i+1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j+1)) / dy;
end
% Diffusive term (central finite difference)
diffusive = D * ((C_temp(i+1,j) - 2*C_temp(i,j) + C_temp(i-1,j)) /
dx^2 + ...
(C_temp(i,j+1) - 2*C_temp(i,j) + C_temp(i,j-1)) /
dy^2);

% Update concentration using implicit Euler's method

C(i,j,n+1) = C_temp(i,j) + dt * (convective_x + convective_y +
diffusive);
end
end

% Apply boundary conditions

% West Boundary (x = 0)
C(1,:,n+1) = cw0(1:Ny);

% East Boundary (x = L)
% Neumann boundary condition: dC/dx = 0
C(Nx,:,n+1) = C(Nx-1,:,n+1);

% South Boundary (y = 0)
% Neumann boundary condition: dC/dy = 0 (implicit)
 C(:,1,n+1) = C(:,2,n+1);

% North Boundary (y = Ly)

C(:,Ny,n+1) = cnL(1:Nx);
end

% Return final concentration

C_final = C(:,:,Nt);
end
Lx=50;
Ly=1;
Nx=100;
Ny=20;
dx = 0.5051; % x-direction mesh size
dy = 0.0526; % y-direction mesh size
dt = 0.1; %% time step size
u = 1.5; % x-velocity
v = 0; % y-velocity
D = 0.001; % Diffusion coefficient
% Solve advection-diffusion problem with Task 3 implemented
C_final = solve_advection_diffusion_task3(dx, dy, dt, u, v, D);
% Plot concentration distribution at final time
[X, Y] = meshgrid(1:size(C_final,2), 1:size(C_final,1));
surf(X, Y, C_final);
xlabel('y');
ylabel('x');
zlabel('Concentration');
title('Concentration Distribution at Final Time (Task 3)');
Results
Task 4:
function [C_at_times] = solve_advection_diffusion_task4(dx, dy, dt, u, v, D, time_instants)
% Define parameters
Nx = round(100 / dx); % Number of grid points in x-direction
Ny = round(1 / dy); % Number of grid points in y-direction
Nt = round(max(time_instants) / dt); % Number of time steps

% Define boundary conditions

cw0 = @(y) zeros(size(y)); % West boundary condition function
cnL = @(x) 0.1 * (x / Nx) .* ones(size(x)); % North boundary condition function

% Initialize concentration matrix

C = zeros(Nx, Ny, Nt);

% Set initial condition

% Assuming initial concentration is 0 everywhere
C(:,:,1) = zeros(Nx, Ny);

% Define Peclet number

Pe_x = u * dt / dx;
Pe_y = v * dt / dy;

% Time-stepping loop
for n = 1:Nt-1
% Internal nodes update (excluding boundary nodes)
C_temp = C(:,:,n);
for i = 2:Nx-1
for j = 2:Ny-1
% Convective term (hybrid scheme)
if abs(Pe_x) < 2 && abs(Pe_y) < 2
convective_x = u * (C_temp(i,j) - C_temp(i-1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j-1)) / dy;
else
convective_x = u * (C_temp(i,j) - C_temp(i+1,j)) / dx;
convective_y = v * (C_temp(i,j) - C_temp(i,j+1)) / dy;
end

% Diffusive term (central finite difference)

diffusive = D * ((C_temp(i+1,j) - 2*C_temp(i,j) + C_temp(i-1,j)) / dx^2 + ...
(C_temp(i,j+1) - 2*C_temp(i,j) + C_temp(i,j-1)) / dy^2);

% Update concentration using implicit Euler's method

C(i,j,n+1) = C_temp(i,j) + dt * (convective_x + convective_y + diffusive);
end
end

% Apply boundary conditions

% West Boundary (x = 0)
C(1,:,n+1) = cw0(1:Ny);

% East Boundary (x = L)
% Neumann boundary condition: dC/dx = 0
C(Nx,:,n+1) = C(Nx-1,:,n+1);

% South Boundary (y = 0)
% Neumann boundary condition: dC/dy = 0 (implicit)
 C(:,1,n+1) = C(:,2,n+1);

% North Boundary (y = H)
C(:,Ny,n+1) = cnL(1:Nx);
end

% Extract concentration at specified time instants

C_at_times = zeros(Nx, Ny, numel(time_instants));
for i = 1:numel(time_instants)
t_index = round(time_instants(i) / dt);
C_at_times(:,:,i) = C(:,:,t_index);
end
end
% Define parameters
dx = 0.5051; % x-direction mesh size
dy = 0.0526; % y-direction mesh size
dt = 0.1; % time step size (s)
u = 1.5; % x-velocity (m/s)
v = 0; % y-velocity (m/s)
D = 0.01; % Diffusion coefficient

% Time instants to visualize (in seconds)

time_instants = [1, 5, 10, 25, 50, 100];

% Solve advection-diffusion equation for Task 4

C_at_times = solve_advection_diffusion_task4(dx, dy, dt, u, v, D, time_instants);

% Plot concentration distribution at different time instants

for i = 1:numel(time_instants)
figure;
surf(C_at_times(:,:,i));
xlabel('y');
ylabel('x');
zlabel('Concentration');
title(['Concentration Distribution at t = ', num2str(time_instants(i)), 's']);
end

% Sensitivity Analysis to Mesh Sizes

mesh_sizes = [0.5, 0.2, 0.1]; % Different mesh sizes to test

for dx_test = mesh_sizes

dy_test = dx_test / 10; % Keeping aspect ratio consistent
C_at_times_mesh = solve_advection_diffusion_task4(dx_test, dy_test, dt, u, v, D, time_instants);
% Plot concentration distribution at t = 50s for each mesh size
figure;
surf(C_at_times_mesh(:,:,end));
xlabel('y');
ylabel('x');
zlabel('Concentration');
title(['Concentration Distribution at t = 50s with Mesh Size dx = ', num2str(dx_test), 'm']);
end

% Sensitivity Analysis to Time Step Sizes

time_step_sizes = [0.5, 0.2, 0.1]; % Different time step sizes to test
for dt_test = time_step_sizes
C_at_times_time = solve_advection_diffusion_task4(dx, dy, dt_test, u, v, D, time_instants);
% Plot concentration distribution at t = 50s for each time step size
figure;
surf(C_at_times_time(:,:,end));
xlabel('y');
ylabel('x');
zlabel('Concentration');
title(['Concentration Distribution at t = 50s with Time Step dt = ', num2str(dt_test), 's']);
end

Figure 1: Concentration Distribution at t=1s Figure 2: Concentration Distribution at t=5s

Figure 3: Concentration Distribution at t=10s Figure 4: Concentration Distribution at t=25s

Figure 5: Concentration Distribution at t=50s Figure 2: Concentration Distribution at t=100s

Sensitivity Analysis to Mesh Size:

Figure 7: Concentration Distribution at t=50s Figure 8: Concentration Distribution at t=50s
With Mesh size dx=0.5m With Mesh size dx=0.2m

Figure 9: Concentration Distribution at t=50s

With Mesh size dx=0.1m
Sensitivity Analysis to Time Step Analysis
Figure 10: Concentration Distribution at t=50s Figure 11: Concentration Distribution at t=50s
With Time step dt=0.5s With Time step dt=0.2s

Figure 12: Concentration Distribution at t=50s

With Time step dt=0.1s
We use a finite difference method to discretize the spatial and temporal derivatives and solve the resulting
system of equations numerically. The central difference method is used for the diffusion term, and a hybrid
scheme is employed for the advection term to ensure stability.
Results:
After implementing the solution approach, we obtained concentration distributions (C) at different time
instants (t). The results illustrate how the concentration evolves over time within the rectangular domain.
Visualizations such as surface plots or contour plots can effectively demonstrate the spatial distribution of
concentration at each time step.
Sensitivity Analysis:
To assess the sensitivity of the results, we conducted two analyses:

• Mesh Size Sensitivity: We varied the mesh sizes (Δx and Δy) and observed their impact on the
concentration distribution at a specific time instant (e.g., t=50s). Changes in mesh sizes can affect
the spatial resolution of the solution and potentially influence the accuracy of the results.
• Time Step Size Sensitivity: Similarly, we varied the time step size (Δt) and analyzed its effect on
the concentration distribution at t=50s. Adjusting the time step size can influence the temporal
resolution of the solution and affect the stability and accuracy of the numerical simulation.
Discussion:
By analyzing the sensitivity of the results to changes in mesh sizes and time step sizes, we gain insights
into the numerical solution's robustness and accuracy. The discussion includes observations on how
different parameter settings impact the concentration distribution and recommendations for selecting
appropriate mesh and time step sizes to achieve accurate and efficient simulations. Additionally, we may
discuss potential limitations or challenges encountered during the sensitivity analysis and propose strategies
to address them.
During the sensitivity analysis of the advection-diffusion problem, we may encounter several limitations
and challenges:

• Computational Resources: Performing sensitivity analysis for different mesh sizes and time step
sizes can be computationally intensive, especially for large domains or fine resolutions. This may
require significant computational resources and time.
• Accuracy vs. Efficiency Trade-off: Using smaller mesh sizes and time step sizes generally
improves the accuracy of the numerical solution but increases computational cost. Finding the
optimal balance between accuracy and computational efficiency is crucial.
• Numerical Stability: As we vary the mesh sizes and time step sizes, the numerical scheme's
stability may be affected. Certain combinations of mesh sizes and time step sizes could lead to
unstable solutions, such as numerical oscillations or divergence. Ensuring numerical stability is
essential for obtaining reliable results.
• Convergence Issues: The iterative solver used to solve the discretized equations may encounter
convergence issues for certain combinations of mesh sizes and time step sizes. Convergence
problems can arise due to rapid changes in the solution or insufficient iterations.
To address these limitations and challenges, we can employ the following strategies:

• Parallel Computing: Utilize parallel computing techniques or distributed computing resources to

accelerate the sensitivity analysis and reduce computational time.
• Adaptive Mesh Refinement: Implement adaptive mesh refinement techniques that dynamically
adjust the mesh sizes based on solution gradients or error estimations. This approach focuses
computational resources on regions where accuracy is most critical.
• Stability Analysis: Perform stability analysis to identify critical stability limits for different mesh
sizes and time step sizes. Ensure that the chosen mesh sizes and time step sizes lie within stable
regions to avoid numerical instability.
• Robust Solvers: Use robust iterative solvers with appropriate convergence criteria to handle
convergence issues efficiently. Adjust solver parameters or switch to alternative solvers if
convergence problems persist.
• Error Estimation: Employ error estimation techniques to quantitatively assess the solution's
accuracy for different mesh sizes and time step sizes. This helps in determining whether further
refinement is necessary and provides confidence in the results.

Conclusion
In conclusion, the problem of advection-diffusion provides valuable insights into the behavior of a scalar
quantity transported by fluid flow and diffusive processes. Throughout the analysis, we tackled four main
tasks:

• Discretization and Implementation: We discretized the advection-diffusion equation using

implicit Euler's method and central finite difference schemes for the diffusive term. The
implementation involved handling boundary conditions and selecting appropriate numerical
schemes based on the Peclet number.
• Boundary Conditions: Neumann boundary conditions were implicitly incorporated into the
governing equations for the east and south boundaries. This ensured that the boundary conditions
were satisfied during the numerical solution process.
• Sensitivity Analysis: We conducted a sensitivity analysis to investigate the impact of mesh sizes
and time step sizes on the numerical solution's accuracy and stability. This involved varying these
parameters and examining their effects on the concentration distribution, particularly at t = 50s.
• Discussion and Analysis: The results of the sensitivity analysis were discussed, highlighting the
trade-offs between accuracy and computational efficiency. We also addressed potential limitations
and challenges encountered during the analysis, proposing strategies to mitigate them.
The analysis provided valuable insights into the behavior of the advection-diffusion problem and
demonstrated the importance of careful parameter selection and numerical methodology in obtaining
reliable solutions.