Skip to content
Permalink
d5ea9d970c
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

ME3255S2017/lecture_07/lecture_07.md

Go to file
 
 
Cannot retrieve contributors at this time
Roots: Open methods Newton-Raphson Secant Methods Modified Secant method Matlab's function Comparison of Solvers
448 lines (272 sloc) 10 KB
Raw Blame 
%plot --format svg
setdefaults

Roots: Open methods

Newton-Raphson

First-order approximation for the location of the root (i.e. assume the slope at the given point is constant, what is the solution when f(x)=0)

$f'(x_{i})=\frac{f(x_{i})-0}{x_{i}-x_{i+1}}$

$x_{i+1}=x_{i}-\frac{f(x_{i})}{f'(x_{i})}$

Use Newton-Raphson to find solution when $e^{-x}=x$

f= @(x) exp(-x)-x;
df= @(x) -exp(-x)-1;

x_i= 0;
x_r = x_i-f(x_i)/df(x_i)
error_approx = abs((x_r-x_i)/x_r)
x_i=x_r;

x_r =  0.50000
error_approx =  1

x_r = x_i-f(x_i)/df(x_i)
error_approx = abs((x_r-x_i)/x_r)
x_i=x_r;
x_r =  0.56631
error_approx =  0.11709

x_r = x_i-f(x_i)/df(x_i)
error_approx = abs((x_r-x_i)/x_r)
x_i=x_r;
x_r =  0.56714
error_approx =  0.0014673

x_r = x_i-f(x_i)/df(x_i)
error_approx = abs((x_r-x_i)/x_r)
x_i=x_r;
x_r =  0.56714
error_approx =    2.2106e-07

In the bungee jumper example, we created a function f(m) that when f(m)=0, then the mass had been chosen such that at t=4 s, the velocity is 36 m/s.

$f(m)=\sqrt{\frac{gm}{c_{d}}}\tanh(\sqrt{\frac{gc_{d}}{m}}t)-v(t)$.

to use the Newton-Raphson method, we need the derivative $\frac{df}{dm}$

$\frac{df}{dm}=\frac{1}{2}\sqrt{\frac{g}{mc_{d}}}\tanh(\sqrt{\frac{gc_{d}}{m}}t)- \frac{g}{2m}\mathrm{sech}^{2}(\sqrt{\frac{gc_{d}}{m}}t)$

setdefaults
g=9.81; % acceleration due to gravity
m=linspace(50, 200,100); % possible values for mass 50 to 200 kg
c_d=0.25; % drag coefficient
t=4; % at time = 4 seconds
v=36; % speed must be 36 m/s
f_m = @(m) sqrt(g*m/c_d).*tanh(sqrt(g*c_d./m)*t)-v; % anonymous function f_m
df_m = @(m) 1/2*sqrt(g./m/c_d).*tanh(sqrt(g*c_d./m)*t)-g/2./m*sech(sqrt(g*c_d./m)*t).^2;
[root,ea,iter]=newtraph(f_m,df_m,140,0.00001)
root =  142.74
ea =    8.0930e-06
iter =  48

Secant Methods

Not always able to evaluate the derivative. Approximation of derivative:

$f'(x_{i})=\frac{f(x_{i-1})-f(x_{i})}{x_{i-1}-x_{i}}$

$x_{i+1}=x_{i}-\frac{f(x_{i})}{f'(x_{i})}$

$x_{i+1}=x_{i}-\frac{f(x_{i})}{\frac{f(x_{i-1})-f(x_{i})}{x_{i-1}-x_{i}}}= x_{i}-\frac{f(x_{i})(x_{i-1}-x_{i})}{f(x_{i-1})-f(x_{i})}$

What values should $x_{i}$ and $x_{i-1}$ take?

To reduce arbitrary selection of variables, use the

Modified Secant method

Change the x evaluations to a perturbation $\delta$.

$x_{i+1}=x_{i}-\frac{f(x_{i})(\delta x_{i})}{f(x_{i}+\delta x_{i})-f(x_{i})}$

[root,ea,iter]=mod_secant(f_m,1,50,0.00001)
root =  142.74
ea =    3.0615e-07
iter =  7

car_payments(400,30000,0.05,5,1)
ans =    1.1185e+04

svg

Amt_numerical=mod_secant(@(A) car_payments(A,700000,0.0875,30,0),1e-6,50,0.001)
car_payments(Amt_numerical,700000,0.0875,30,1)
Amt_numerical =  5467.0
ans =    3.9755e-04

svg

Amt_numerical*12*30
ans =    1.9681e+06

Amortization calculation makes the same calculation for the monthly payment amount, A, paying off the principle amount, P, over n pay periods with monthly interest rate, r.

% Amortization calculation
A  = @(P,r,n) P*(r*(1+r)^n)./((1+r)^n-1);
Amt=A(30000,0.05/12,5*12)
Amt =  566.14

Matlab's function

Matlab and Octave combine bracketing and open methods in the fzero function.

help fzero
'fzero' is a function from the file /usr/share/octave/4.0.0/m/optimization/fzero.m

 -- Function File: fzero (FUN, X0)
 -- Function File: fzero (FUN, X0, OPTIONS)
 -- Function File: [X, FVAL, INFO, OUTPUT] = fzero (...)
     Find a zero of a univariate function.

     FUN is a function handle, inline function, or string containing the
     name of the function to evaluate.

     X0 should be a two-element vector specifying two points which
     bracket a zero.  In other words, there must be a change in sign of
     the function between X0(1) and X0(2).  More mathematically, the
     following must hold

          sign (FUN(X0(1))) * sign (FUN(X0(2))) <= 0

     If X0 is a single scalar then several nearby and distant values are
     probed in an attempt to obtain a valid bracketing.  If this is not
     successful, the function fails.

     OPTIONS is a structure specifying additional options.  Currently,
     'fzero' recognizes these options: "FunValCheck", "OutputFcn",
     "TolX", "MaxIter", "MaxFunEvals".  For a description of these
     options, see *note optimset: XREFoptimset.

     On exit, the function returns X, the approximate zero point and
     FVAL, the function value thereof.

     INFO is an exit flag that can have these values:

        * 1 The algorithm converged to a solution.

        * 0 Maximum number of iterations or function evaluations has
          been reached.

        * -1 The algorithm has been terminated from user output
          function.

        * -5 The algorithm may have converged to a singular point.

     OUTPUT is a structure containing runtime information about the
     'fzero' algorithm.  Fields in the structure are:

        * iterations Number of iterations through loop.

        * nfev Number of function evaluations.

        * bracketx A two-element vector with the final bracketing of the
          zero along the x-axis.

        * brackety A two-element vector with the final bracketing of the
          zero along the y-axis.

     See also: optimset, fsolve.

Additional help for built-in functions and operators is
available in the online version of the manual.  Use the command
'doc <topic>' to search the manual index.

Help and information about Octave is also available on the WWW
at http://www.octave.org and via the help@octave.org
mailing list.

fzero(@(A) car_payments(A,30000,0.05,5,0),500)
ans =  563.79

Comparison of Solvers

It's helpful to compare to the convergence of different routines to see how quickly you find a solution.

Comparing the freefall example

N=20;
iterations = linspace(1,400,N);
ea_nr=zeros(1,N); % appr error Newton-Raphson
ea_ms=zeros(1,N); % appr error Modified Secant
ea_fp=zeros(1,N); % appr error false point method
ea_bs=zeros(1,N); % appr error bisect method
for i=1:length(iterations)
    [root_nr,ea_nr(i),iter_nr]=newtraph(f_m,df_m,300,0,iterations(i));
    [root_ms,ea_ms(i),iter_ms]=mod_secant(f_m,1e-6,300,0,iterations(i));
    [root_fp,ea_fp(i),iter_fp]=falsepos(f_m,1,300,0,iterations(i));
    [root_bs,ea_bs(i),iter_bs]=bisect(f_m,1,300,0,iterations(i));
end

setdefaults
semilogy(iterations,abs(ea_nr),iterations,abs(ea_ms),iterations,abs(ea_fp),iterations,abs(ea_bs))
legend('newton-raphson','mod-secant','false point','bisection')
warning: axis: omitting non-positive data in log plot
warning: called from
    __line__ at line 120 column 16
    line at line 56 column 8
    __plt__>__plt2vv__ at line 500 column 10
    __plt__>__plt2__ at line 246 column 14
    __plt__ at line 133 column 15
    semilogy at line 60 column 10
warning: axis: omitting non-positive data in log plot
warning: axis: omitting non-positive data in log plot
warning: axis: omitting non-positive data in log plot

svg

ea_nr
ea_nr =

 Columns 1 through 8:

   6.36591   0.06436   0.00052   0.00000   0.00000   0.00000   0.00000   0.00000

 Columns 9 through 16:

   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000   0.00000

 Columns 17 through 20:

   0.00000   0.00000   0.00000   0.00000

N=20;
f= @(x) x^10-1;
df=@(x) 10*x^9;
iterations = linspace(1,50,N);
ea_nr=zeros(1,N); % appr error Newton-Raphson
ea_ms=zeros(1,N); % appr error Modified Secant
ea_fp=zeros(1,N); % appr error false point method
ea_bs=zeros(1,N); % appr error bisect method
for i=1:length(iterations)
    [root_nr,ea_nr(i),iter_nr]=newtraph(f,df,0.5,0,iterations(i));
    [root_ms,ea_ms(i),iter_ms]=mod_secant(f,1e-6,0.5,0,iterations(i));
    [root_fp,ea_fp(i),iter_fp]=falsepos(f,0,5,0,iterations(i));
    [root_bs,ea_bs(i),iter_bs]=bisect(f,0,5,0,iterations(i));
end
        
semilogy(iterations,abs(ea_nr),iterations,abs(ea_ms),iterations,abs(ea_fp),iterations,abs(ea_bs))
legend('newton-raphson','mod-secant','false point','bisection')
warning: axis: omitting non-positive data in log plot
warning: called from
    __line__ at line 120 column 16
    line at line 56 column 8
    __plt__>__plt2vv__ at line 500 column 10
    __plt__>__plt2__ at line 246 column 14
    __plt__ at line 133 column 15
    semilogy at line 60 column 10
warning: axis: omitting non-positive data in log plot
warning: axis: omitting non-positive data in log plot
warning: axis: omitting non-positive data in log plot

svg

ea_nr
newtraph(f,df,0.5,0,12)
ea_nr =

 Columns 1 through 7:

   99.03195   11.11111   11.11111   11.11111   11.11111   11.11111   11.11111

 Columns 8 through 14:

   11.11111   11.11111   11.11111   11.11109   11.11052   11.10624   10.99684

 Columns 15 through 20:

    8.76956    2.12993    0.00000    0.00000    0.00000    0.00000

ans =  16.208

df(300)
ans =    1.9683e+23

% our class function
f= @(x) tan(x)-(x-1).^2
mod_secant(f,1e-3,1)
f =

@(x) tan (x) - (x - 1) .^ 2

ans =  0.37375

f(ans)
ans =   -3.5577e-13

tan(0.37375)
(0.37375-1)^2
ans =  0.39218
ans =  0.39219

f([0:10])
ans =

 Columns 1 through 8:

   -1.0000    1.5574   -3.1850   -4.1425   -7.8422  -19.3805  -25.2910  -35.1286

 Columns 9 through 11:

  -55.7997  -64.4523  -80.3516