README.md

ME3255_Final_Project

Part A

Problem

Create a central finite difference approximation function to solve for the membrane displacement gradient with 3-by-3 interior nodes '[w]=membrane_solution3(T,P)'. The inputs are tension (T) and pressure (P), and the output is column vector w.

function [w] = membrane_solution3(T,P)
% membrane_solution3: dispalacement of node for membrane with 3x3 interior
% nodes
% [w] = membrane_solution3(T,P)
% input:
% T = Tension (microNewton/micrometer)
% P = Pressure (MPa)
% output:
% w = vector of displacement of interior nodes

od = ones(8,1);
od(3:3:end) = 0;
k = -4*diag(ones(9,1))+diag(ones(9-3,1),3)+diag(ones(9-3,1),-3)+diag(od,1)+diag(od,-1);


y = -(10/4)^2*(P/T)*ones(9,1);
w = k\y;

% Solves for displacement (micrometers)
% Solution represents a 2D data set w(x,y)


    [x,y] = meshgrid(0:10/4:10,0:10/4:10);
    z = zeros(size(x));
    z(2:end-1,2:end-1) = reshape(w,[3 3]);
    surf(x,y,z)
    title('Membrane Displacement (3 x 3)')
    zlabel('Z Position (micron)')
    ylabel('Y Position (micron)')
    xlabel('X Position (micron)')

% Membrane displacement is shown on chart

end
    
end 

Approach

• A matrix is set up using finite element analysis of the interior nodes of a 5x5 matrix
• A vector for the y direction is then set up for the membrane
• Using linear algebra, a vector for the displacement of the nodes is created
• The vector can then be transformed to represent the actual surface as a matrix

Part B

Problem

Use membranesolution3 function to solve for w and plot the result with surf(X,Y,W). Pressure P=0.001 MPa and tension T=0.006 uN/um.

% Part B Script
% From problem statement:
% T=0.006 uN/um
% P=0.001 MPa
[w] = membrane_solution3(0.006,0.001);

Part C

Problem

Create a central finite difference approximation function to solve for the membrane displacement gradient with n-by-n interior nodes '[w]=membrane_solution(T,P,n)'. The inputs are tension (T), pressure (P), and number of nodes (n), and the output is column vector w.

function [w] = membrane_solution(T,P,n)
% membrane_solution: dispalacement of node for membrane with nxn interior nodes
% [w] = membrane_solution(T,P,n)
% input:
% T = Tension (microNewton/micrometer)
% P = Pressure (MPa)
% n = number of rows and columns of interior nodes
% output:
% w = vector of displacement of interior nodes

    od = ones(n^2-1,1);
    od(n:n:end) = 0;
    k = -4*diag(ones(n^2,1))+diag(ones((n^2)-n,1),n)+diag(ones((n^2)-n,1),-n)+diag(od,1)+diag(od,-1);

    y = -(10/(n+1))^2*(P/T)*ones(n^2,1);
    w = k\y;
    % Solves for displacement (micrometers)
    % Output w is a vector
    % Solution represents a 2D data set w(x,y)

    [x,y] = meshgrid(0:10/(n+1):10,0:10/(n+1):10);
    z = zeros(size(x));
    z(2:end-1,2:end-1) = reshape(w,[n n]);
    surf(x,y,z)
    title('Membrane Displacement')
    zlabel('Displacement (micrometer)')
    % Membrane displacement is shown on chart
end

Approach

• This problem uses the same steps as with the membrane_solution3 function, except it allows for the user to change the number of interior nodes
• This required alterations of the all the indexing and sizing in the function along with a few calculations

Part D

Problem

Use membranesolution function to solve for w and plot the result with surf(X,Y,W). Pressure P=0.001 MPa, tension T=0.006 uN/um, and number of nodes n=10.

% Part D Script
% From problem statement:
% T=0.006 uN/um
% P=0.001 MPa
% n = 10 nodes
[w] = membrane_solution(0.006,0.001,10)

Part E

Problem

Create a function that calculates the difference in strain energy and work done by pressure for n-by-n elements '[pw_se,w]=SE_diff(T,P,n)'.

function [pw_se,w]=SE_diff(T,P,n)
% SE_diff: calculates difference between strain energy and work done by pressure in
% membrane
% [pw_se,w]=SE_diff(T,P,n)
% input:
% T = Tension (microNewton/micrometer)
% P = Pressure (MPa)
% n = number of rows and columns of interior nodes
% output:
% pw_se = difference between strain energy and work done by pressure in
% membrane
% w = vector of displacement of interior nodes

E = 1e6; %MPa Units may need to be changed
v = .31; %Poissons ratio
t = .0003; %um
h = 10/(n+1); %um
w = membrane_solution(T,P,n);
z = zeros(n+2);
z(2:end-1,2:end-1) = reshape(w,[n n]);
num = n + 1;
wbar = zeros(num);
for i = 1:num
    for j = 1:num
        wbar(i,j) = mean([z(i,j),z(i+1,j),z(i,j+1),z(i+1,j+1)]);
    end
end
pw = sum(sum(wbar.*h^2.*P));
dwdx = zeros(num);
dwdy = zeros(num);
for i = 1:num
    for j = 1:num
        dwdx(i,j) = mean([z(i+1,j)-z(i,j),z(i+1,j+1)-z(i,j+1)]);
        dwdy(i,j) = mean([z(i,j+1)-z(i,j),z(i+1,j+1)-z(i+1,j)]);
    end
end
se = E*t*h^2/(2*(1-v^2))*sum(sum(0.25.*dwdx.^4+.25.*dwdy.^4+0.5.*(dwdx.*dwdy).^2));
pw_se = pw-se;

Approach

• Using the membrane_solution function, a vector of the displacements is formed
• Next, the average displacement for each element is calculated. For each elements, the dispalcement at all four courners is taken and then averaged
• Using these values, the work done by pressure can be calculated
• For the change in dispalcement over the x and y coordinate system, the values of the change on the left and right (y-axis) or top and bottom (x-axis) are taken and averaged
• Using these values, the strain energy can be calculated

Part F

Problem

Use a root-finding method to calculate the tension in the membrane given a pressure, P=0.001 MPa, and n=[20:5:40] interior nodes. Show the decrease in tension error in a table.

% Part F script
% From Problem Statement:
 n=[3,20:5:40];
 P=0.001; %MPa
 
 % Sets vector length so it doesn't change every iteration in for loop
 T = zeros(1,length(n));
 ea = zeros(1,length(n));
 
 % Uses tension_sol function to solve for the tension for each input
 for i = 1:length(n)
[T(i), ea(i)] = tension_sol(P,n(i));
 end

function [T,ea] = tension_sol(P,n)
% tension_sol: outputs tension of a membrane given the pressure and number
% of nodes
% [T,ea] = tension_sol(P,n)
% input:
% P = Pressure (MPa)
% n = number of rows and columns of interior nodes
% output:
% T = Tension (microNewton/micrometer)
% ea = approximate relative error (%)

y =@(T) SE_diff(T,P,n);
[T,fx,ea,iter]=bisect(y,.01,1);

function [root,fx,ea,iter]=bisect(func,xl,xu,es,maxit,varargin)
% bisect: root location zeroes
% [root,fx,ea,iter]=bisect(func,xl,xu,es,maxit,p1,p2,...):
% uses bisection method to find the root of func
% input:
% func = name of function
% xl, xu = lower and upper guesses
% es = desired relative error (default = 0.0001%)
% maxit = maximum allowable iterations (default = 50)
% p1,p2,... = additional parameters used by func
% output:
% root = real root
% fx = function value at root
% ea = approximate relative error (%)
% iter = number of iterations
if nargin<3,error('at least 3 input arguments required'),end
test = func(xl,varargin{:})*func(xu,varargin{:});
if test>0,error('no sign change'),end
if nargin<4||isempty(es), es=0.0001;end
if nargin<5||isempty(maxit), maxit=50;end
iter = 0; xr = xl; ea = 100;
while (1)
  xrold = xr;
  xr = (xl + xu)/2;
  iter = iter + 1;
  if xr ~= 0,ea = abs((xr - xrold)/xr) * 100;end
  test = func(xl,varargin{:})*func(xr,varargin{:});
  if test < 0
    xu = xr;
  elseif test > 0
    xl = xr;
  else
    ea = 0;
  end
  if ea <= es || iter >= maxit,break,end
end
root = xr; fx = func(xr, varargin{:});

function re = Rel_error (T)
 Rel_error: calculates relative error of a vector
% re = Rel_error (T)
% input:
% T = vector of numbers
% output:
% re = relative error of vector

re = zeros(1,length(T)-1);
for i = 2:length(T)
    re(i-1)= abs(T(i)-T(i-1))/T(i-1);
end

number of nodes	Tension (uN/um)	rel. error
3	0.0489	n/a
20	0.0599	22.6%
25	0.0601	0.27%
30	0.0602	0.15%
35	0.0602	0.09%
40	0.0603	0.06%

Approach

• This problem uses the bisect method for locating roots and the SE_diff function for calculating the difference in work and strain energy of the membrane
• The script runs through all iterations of different amounts of nodes, zeroing the SE_diff function output and saving the values for tension in the T variable as a vector

Part G

Problem

Plot the pressure vs maximum deflection for pressure range 0.001-0.01 MPa with 10 steps using a root-finding method to determine tension, T, at each pressure.

% Part G Script
% From Problem Statement
P = linspace(.001,.01,10);
n = 20;

% Sets vector length so it doesn't change every iteration in for loop
T = zeros(1,length(P));
wmax = zeros(1,length(P));

% Uses tension_sol and membrane_solution functions to solve for the maximum displacement for all iterations
for i = 1:length(P)
    T(i) = tension_sol(P(i),n);
    w = membrane_solution(T(i),P(i),n);
    wmax(i) = max(w);
end

% Sets up figure for plot
clf
setDefaults

% Standard Linear Regression with equation P(x) = A*w^2
x = wmax';
y = P';
Z=x.^3;
a=Z\y;
x_fcn=linspace(min(x),max(x));

% Plots regression line with actual points
plot(x,y,'o',x_fcn,a*x_fcn.^3)
title('Pressure vs Maximum Deflection')
xlabel('Maximum Deflection (um)')
ylabel('Pressure (MPa)')
print('Part g','-dpng')

Approach

• This script uses the tension_sol function as described in the part above, running though all iterations of pressure
• Additionally, the max deflection for each pressure and tension is calculated at each iteration
• From there, a general linear regression is calculated with the formula P(x) = A*w^3
• The results are plotted

Part H

Problem

Show that the constant A is converging as the number of nodes is increased.

% Part G Script
% From Problem Statement
P=0.001;
n = 5:5:50;

% Sets vector length so it doesn't change every iteration in for loop
a = zeros(1,length(n));

% Uses tension_sol and membrane_solution functions along with a regression line to solve for the constant A in the regression line
for i = 1:length(n)
    T = tension_sol(P,n(i));
    w = membrane_solution(T,P,n(i));
    wmax = max(w);
    x = wmax';
    y = P';
    Z = x.^3;
    a(i) = Z\y;
end

% Calculates the relative error between the constant A values
Rel_error(a)

number of nodes	vaalue of const. A	rel. error
5	0.4423	n/a
10	0.5389	21.84%
15	0.5349	0.73%
20	0.5483	2.5%
25	0.5454	0.52%
30	0.5503	0.89%
35	0.5485	0.31%
40	0.5510	0.45%
45	0.5499	0.2%

Approach

• Part H uses the same basis as Part G, but with varrying n (number of nodes) valeus instead
• Additionally, the outputs and a few calculations were deleted to speed up the process since they were not needed here
• In the chart, while there are some fluctuations, there is a clear decrease in the relative error of A, leading to the assumption that it is converging as the number of nodes increases