Fractional Calculus George

Diskussion
Fraktionale Differentialgleichungen
Seite 20/31
(d) Using modified Fourier Series

The modification has been done as follows:
Firstly, the Fourier Series approximation of the original signal is solved using exactly
the previous method.
After that, the Fourier Series approximation was filtered to remove the extreme values. A threthold has been taken as follows:
v
u n
u1 X
RM SE = t
e2 ,
n
t=1
et = ReconstructedF unction OriginalF unction

If the individual error is larger than RM2SE , the time instant which causes the error
will be replaced with the coresponding time instant of the original signal.
After that, the filtered Fourier Series coefficients were again reconstructed from the
filtered signal and were used to find the required partial derivative using the Eq (1)
of the previous section.
Matlab Function:
1
2
3
4
5
6
7
8
10
11
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%INPUT
%----% - f : The required function. f must be defined for [t0,t0+T]
% - T
: period of the function
% - M
: number of frequencies (default: 50)
%-------------------------------------------------------------------------function .
[ReturnedValue1,ReturnedValue2,ReturnedValue6,ReturnedValue3,ReturnedValue4,
...
ReturnedValue7,ReturnedValue5] = .
Fractional_Derivative_Fourier_Series_Complete ...
(FunctionToBeDerivedInStringX, Beginning_Point, End_Point, .
RequiredDerivative)
%--------------------------------------------------------------------------
12
13
14
15
16
17
18
19
20
21
22
%-------------------------------------------------------------------------N = 100; % Number of points per period

M = 50; % Should be half of N
InputFunction = FunctionToBeDerivedInStringX;
t0 = Beginning_Point;
T = End_Point - Beginning_Point;
q = RequiredDerivative;
w0 = 2*pi/T;
t = linspace(t0,T,N);
%--------------------------------------------------------------------------
The code of this function was inspired/excerpted/modified by/from both Guilherme Coco fourier_coeff Beltramini and George Papazafeiropoulos fourier_diffint
Diskussion
Seite 21/31
23
24
25
26
27
%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Another Way %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

fun = inline(InputFunction,x);
U = fun(t);
28
29
A0_New = 1/T*quadl(fun,t0,t0+T);
30
31
32
coeff_New(2*M+1,1) = 0;
coeff_New(1) = A0_New;
% Initialization
33
34
35
36
37
f_aux = inline([( InputFunction ) .*cos(w0*m*x)],x,m,w0);

for m=1:M
coeff_New(m+1) = quadl(f_aux,t0,t0+T,[],[],m,w0);
end
38
39
40
41
42
43
f_aux = inline([( InputFunction ) .*sin(w0*m*x)],x,m,w0);

for m=1:M
coeff_New(m+M+1) = quadl(f_aux,t0,t0+T,[],[],m,w0);
end
coeff_New(2:end) = 2/T * coeff_New(2:end);
44
45
%--------------------------------------------------------------------------
46
47
%%%%%%%%%%%%%%% Reconstruct the Signal %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48
49
50
51
52
53
54
55
56
Fourier_Series = zeros(N,2*M+1);
% Initialization
Fourier_Series(:,1) = 1;
for m=2:M+1
Fourier_Series(:,m) = cos(w0*(m-1)*t);
Fourier_Series(:,m+M) = sin(w0*(m-1)*t);
end
Function_Reconstructed = Fourier_Series*reshape(coeff_New,2*M+1,1);
ReturnedValue1 = Function_Reconstructed;
57
58
59
60
%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%%%%%% Perform filteration of extreme values %%%%%%%%%%%%%%

%--------------------------------------------------------------------------
61
62
Error_Method2 = sqrt(mean((Function_Reconstructed - U).^2));
63
64
ErrorVector_Method2 = zeros(1, length(t));
65
66
for i = 1 : length(t)
67
ErrorVector_Method2(i) = sqrt((Function_Reconstructed(i) - U(i)).^2);
68
69
end
70
71
72
73
74
Function_Reconstructed_filtered = Function_Reconstructed;
Indecies1 = find(ErrorVector_Method2 > Error_Method2/2);
Function_Reconstructed_filtered(Indecies1) = U(Indecies1);
75
76
77
Error_Method2_New = sqrt(mean((Function_Reconstructed_filtered - U).^2));

ReturnedValue2 = Error_Method2_New;
Diskussion
78
Seite 22/31
ReturnedValue6 = ErrorVector_Method2;
79
80
81
82
%-------------------------------------------------------------------------%%%%%%%%%%%% Recover Fourier Coefficients of Filtered Signal %%%%%%%%%%%%%%

%--------------------------------------------------------------------------
83
84
Fourier_Series_Method2 = Fourier_Series \ Function_Reconstructed_filtered;
85
86
87
88
89
A0_New_filtered = Fourier_Series_Method2(1);
An_New_filtered = Fourier_Series_Method2(2 : M+1);
Bn_New_filtered = Fourier_Series_Method2(M+2 : 2*M+1);
coeff_New_filtered = [A0_New_filtered ; An_New_filtered ; Bn_New_filtered];
90
91
92
%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%% Reconstruct filted signals %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
93
94
95
96
97
98
99
Fourier_Series_filtered = zeros(N,2*M+1);
% Initialization
Fourier_Series_filtered(:,1) = 1;
for m=2:M+1
Fourier_Series_filtered(:,m) = cos(w0*(m-1)*t);
Fourier_Series_filtered(:,m+M) = sin(w0*(m-1)*t);
end
100
101
102
Function_Reconstructed_filtered_reconstruct = .
Fourier_Series_filtered*reshape(coeff_New_filtered,2*M+1,1);
ReturnedValue3 = Function_Reconstructed_filtered_reconstruct;
103
104
ErrorVector_Method_Reconstructed = zeros(1, length(t));
105
106
for i = 1 : length(t)
107
ErrorVector_Method_Reconstructed(i) = .
sqrt((Function_Reconstructed_filtered_reconstruct(i) - U(i)).^2);
108
109
end
110
111
112
113
Reconstruction_Error_Method2 = .
sqrt(mean((Function_Reconstructed_filtered_reconstruct - U).^2));
ReturnedValue4 = Reconstruction_Error_Method2;
ReturnedValue7 = ErrorVector_Method_Reconstructed;
114
115
116
%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%%%%% Calculate the Partial Derivative %%%%%%%%%%%%%%%%%%%%
117
118
A0_Fourier = A0_New_filtered/(gamma(1 - q) + eps);
119
120
121
122
123
124
125
126
127
128
129
n = length(An_New_filtered);
KK = (1:n) * w0;
ZZ = (q * pi)/2;
% replicate data appropriately
eval_t = t(:,ones(1,n));
eval_n = KK(:,ones(1,length(t)));
eval_An = An_New_filtered(:,ones(1,length(t)));
eval_Bn = Bn_New_filtered(:,ones(1,length(t)));
Diskussion
130
131
Seite 23/31
% differintegral
Y2_Der = A0_Fourier .* (t + eps).^(-q) + sum((eval_n + .
eps).^(q).*(eval_An.*cos(eval_n.*eval_t + ZZ) + eval_Bn.*sin(eval_n.*eval_t
+ ZZ)),2);
132
133
134
ReturnedValue5 = Y2_Der;
%--------------------------------------------------------------------------
Example (1): f(x) = sin(x)

1
2
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% This example applies the modified Fourier Series and Then derives it
3
4
5
6
7
8
9
10
11
12
clear all;clc;
figure
t = linspace(0,2*pi,100);
InputFunction = sin(x);
U = fun(t);
alphaX = -10 : 1 : 10;
cc = jet(length(alphaX));
N = 0;
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
for alpha = -10 : 1 : 10

[Y1,Error_Between_Filted_And_Original,Error_Filted_And_Original_Vector, ...
Y3,Error_Recons_Signal_After_Filterat_And_Original, ...
Error_Recons_Signal_After_Filterat_And_Original_Vector,Y5] = ...
Fractional_Derivative_Fourier_Series_Complete(InputFunction, 0, 2*pi, .
alpha/10);
if(length(Y5) == 1)
Y5(1:length(X)) = Y5(length(Y5));
else
%do nothing
end
if(alpha == -10 || alpha == 0 || alpha == 10)
N = 2.5;
else
N = 1.5;
end
plot(t,Y5,LineWidth,N,color,cc(alpha+11,:));hold on
end
32
33
34
set(gca,xTick,0:pi/2:2*pi)
set(gca,xTickLabel,{0,\pi/2,\pi,3\pi/2,2\pi})
35
36
37
38
xlabel(x,fontsize,14,Interpreter,latex);
ylabel($ y = d^{\alpha} f(x) / d .
x^{\alpha}$,fontsize,14,Interpreter,latex);
title(Fractional derivatives of $f(x) = sin(x)$ of order $\alpha = 0, 0.1,
..., 0.2, 1.0$,fontsize,14,Interpreter,latex)
39
40
41
h = legend($\alpha = -1$, $\alpha = -0.9$, $\alpha = -0.8$, $\alpha =
Diskussion
Seite 24/31
-0.7$, $\alpha = -0.6$, $\alpha = -0.5$, $\alpha = -0.4$, $\alpha =

-0.3$, $\alpha = -0.2$, $\alpha = -0.1$, ...
$\alpha = 0$, $\alpha = 0.1$, $\alpha = 0.2$, $\alpha = 0.3$, .
$\alpha = 0.4$, $\alpha = 0.5$, $\alpha = 0.6$, $\alpha = 0.7$,
$\alpha = 0.8$, $\alpha = 0.9$, $\alpha = 1$);
set(h,Interpreter,latex);grid on
42
43
44
45
46
47
text(1.57,-0.4,$y = .
cos(x)$,fontsize,14,color,cc(10+11,:),Interpreter,latex)
text(3.35,0.8,$y = .
-cos(x)$,fontsize,14,color,cc(-10+11,:),Interpreter,latex)
text(3.35,0.2,$y = .
sin(x)$,fontsize,14,color,cc(0+11,:),Interpreter,latex)
48
49
50
51
52
53
54
55
figure;
plot(t,U,LineWidth,2)
hold on
plot(t,Y3,r.,LineWidth,2)
grid on
56
57
58
59
60
61
ylabel($ y = sin(x)$,fontsize,14,Interpreter,latex);
title(Fourier Series approximation of $f(x) = sin(x)$, Extreme values are
filtered,fontsize,14,Interpreter,latex)
h2 = legend($Original$, $Approximation$);
set(h2,Interpreter,latex);
62
63
64
65
66
67
figure;
plot(t,Error_Filted_And_Original_Vector,LineWidth,2)
hold on
plot(t,Error_Recons_Signal_After_Filterat_And_Original_Vector,r.,LineWidth,2)
grid on
68
69
70
71
72
73
ylabel($\sqrt{\frac{1}{n}\sum_{t=1}^{n}e_t^2}$,fontsize,14,Interpreter,latex);
title(Root mean squared Error between Original signal and Fourier Series .
approximation,fontsize,14,Interpreter,latex)
h3 = legend($Before Filtration$, $After Filtration$);
Diskussion
Fractional derivatives of f (x) = sin(x) of order = 1, 0.9, ..., 0.9, 1.0
1.5
y = d f (x)/dx
Seite 25/31
= 1
= 0.9
= 0.8
= 0.7
= 0.6
= 0.5
= 0.4
= 0.3
= 0.2
= 0.1
=0
= 0.1
= 0.2
= 0.3
= 0.4
= 0.5
= 0.6
= 0.7
= 0.8
= 0.9
=1
y = cos(x)
0.5
y = sin(x)
0
0.5
y = cos(x)
1
1.5
3
2
Abbildung 7: Fractional Derivative Calculation using modified Fourier Series. Very powerful method, works for all R. Works perfectly for periodic functions, however, within non periodic
functions Gibbs phenomenon will make the result little poor!
Fourier Series approximation of f (x) = sin(x), Extreme values are filtered
Root mean
squared Error between Original signal and Fourier Series approximation
109
1.8
1
Original
Fourier Series Approximation
0.8
Before Filtration
After Filtration
1.6
0.6
1.4
0.4
2
t=1 et
0.8
q P
n
1
y = sin(x)
1.2
0.2
0
0.2
0.6
0.4
0.4
0.6
0.2
0.8
1
3
2
3
2
Abbildung 8: Modified Fourier Series approximation of f (x) = sin(x) where the extreme values
are filtered based on Root Mean Squared Error threshold.
Diskussion
Seite 26/31
Example (2): f(x) = x*(x pi)*(x 2*pi)

1
2
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3
4
5
6
7
8
9
10
11
12
clear all;clc;
figure
InputFunction = x.*(x - pi).*(x - 2*pi);
U = fun(t);
alphaX = -10 : 1 : 10;
N = 0;
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
%%
for alpha = -10 : 1 : 10
Error_Recons_Signal_After_Filterat_And_Original_Vector,Y5, ...
Threshold] = Fractional_Derivative_Fourier_Series_Complete ...
(InputFunction, 0, 2*pi, alpha/10);
if(length(Y5) == 1)
Y5(1:length(t)) = Y5(length(Y5));
else
%do nothing
end
N = 2.5;
else
N = 1.5;
end
end
%%
36
37
38
39
40
ylabel($ y = d^{\alpha} f(x) / d x^{\alpha}$,fontsize,14,Interpreter, ...
latex);
title(Fractional derivatives of $f(x) = x*(x - pi)*(x - 2*pi)$ of order .
$\alpha = -1, -0.9, ..., 0.9, 1.0$,fontsize,14,Interpreter,latex)
41
42
43
44
45

-0.3$, $\alpha = -0.2$, $\alpha = -0.1$, ...
46
47
% text(1.57,-0.4,$y =
.
.
Diskussion
Seite 27/31
% text(3.35,0.8,$y = .
% text(3.35,0.2,$y = .
48
49
50
51
52
53
54
55
56
57
figure;
plot(t,U,LineWidth,1.5)
hold on
grid on
58
59
60
61
62
63
ylabel($ f(x) = x*(x - pi)*(x - 2*pi)$,fontsize,14,Interpreter,latex);
title(Fourier Series approximation. Extreme values are .
64
65
66
67
68
69
70
71
figure;
plot(t,Error_Filted_And_Original_Vector,LineWidth,1.5)
hold on
hold on
Threshold(1:length(t)) = Threshold(length(Threshold));
plot(t,Threshold,g-,LineWidth,1);grid on
72
73
74
75
76
77
1
2
3
4
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
syms x;
Y = x.*(x - pi).*(x - 2*pi);
t = linspace(0,2*pi,100); % As defined in the fractional calculus function
5
6
7
8
%------ alpha = 0, 0th fractional derivative -----------------------------U = subs(Y,x,t);

plot(t,U,b.); hold on
9
10
11
12
13
%------ alpha = 1, 1st fractional derivative / normal derivative ---------R = diff(Y,x);

R_X = subs(R,x,t);
plot(t,R_X,b.); hold on
14
15
16
17
18
%----- alpha = -1, -1st fractional derivative / normal integration -------RR = int(Y,x);
RR_X = subs(RR,x,t);
plot(t,RR_X,r.); grid on
Diskussion
Seite 28/31
Fractional derivatives of f (x) = x (x pi) (x 2 pi) of order = 1, 0.9, ..., 0.9, 1.0
25
= 1
= 0.9
= 0.8
= 0.7
= 0.6
= 0.5
= 0.4
= 0.3
= 0.2
= 0.1
=0
= 0.1
= 0.2
= 0.3
= 0.4
= 0.5
= 0.6
= 0.7
= 0.8
= 0.9
=1
20
Seems to be
shifted above!
y = d f (x)/dx
15
10
0
5
10
15
3
2
Abbildung 9: Fractional Derivative Calculation using modified Fourier Series.

Fourier Series approximation. Extreme values are filtered
Root mean
squared Error between Original signal and Fourier Series approximation
103
1
15
Original
Fourier Series Approximation
Before Filtration
After Filtration
Threshold
10
0.8
0.7
2
t=1 et
0.6
q P
n
1
0.5
f (x) = x (x pi) (x 2 pi)
0.9
0.4
0.3
0.2
10
15
0.1
0
3
2
3
2
Abbildung 10: Modified Fourier Series approximation of f (x) = x (x pi) (x 2 pi) where
the extreme values are filtered based on Root Mean Squared Error threshold.
Diskussion
Seite 29/31
Example (3): f(x) = x

1
2
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%% My Fourier Series Function With Enhancement & Filteration %%%%%%%
3
4
5
6
7
8
9
10
11
12
13
14
15
tic
clear all;clc;
figure
InputFunction = x;
U = fun(t);
alphaX = -10 : 1 : 10;
N = 0;
16
17
18
19
20
21
22
23
%%
for alpha = -10 : 1 : 10
Error_Recons_Signal_After_Filterat_And_Original_Vector,Y5, ...
Threshold] = Fractional_Derivative_Fourier_Series_Complete ...
(InputFunction, 0, 2*pi, alpha/10);
24
if(length(Y5) == 1)
Y5(1:length(t)) = Y5(length(Y5));
else
%do nothing
end
N = 2.5;
else
N = 1.5;
end
25
26
27
28
29
30
31
32
33
34
35
36
37
end
%%
38
39
40
41
42
43
44
ylabel($ y = d^{\alpha} f(x) / d .
x^{\alpha}$,fontsize,14,Interpreter,latex);
title(Fractional derivatives of $f(x) = x$ of order $\alpha = -1, -0.9, ...,
0.9, 1.0$,fontsize,14,Interpreter,latex)
45
46
47
48

-0.3$, $\alpha = -0.2$, $\alpha = -0.1$, ...
.
.
Diskussion
Seite 30/31

49
50
51
% text(1.57,-0.4,$y = .
% text(3.35,0.8,$y = .
% text(3.35,0.2,$y = .
52
53
54
55
56
57
58
59
60
61
figure;
plot(t,U,LineWidth,1.5)
hold on
grid on
62
63
64
65
66
67
ylabel($ f(x) = x$,fontsize,14,Interpreter,latex);
title(Fourier Series approximation. Extreme values are .
68
69
70
71
72
73
74
75
figure;
plot(t,Error_Filted_And_Original_Vector,LineWidth,1.5)
hold on
hold on
Threshold(1:length(t)) = Threshold(length(Threshold));
plot(t,Threshold,g-,LineWidth,1);grid on
76
77
78
79
80
81
82
toc
Diskussion
Seite 31/31
Fractional derivatives of f (x) = x of order = 1, 0.9, ..., 0.9, 1.0
25
= 1
= 0.9
= 0.8
= 0.7
= 0.6
= 0.5
= 0.4
= 0.3
= 0.2
= 0.1
=0
= 0.1
= 0.2
= 0.3
= 0.4
= 0.5
= 0.6
= 0.7
= 0.8
= 0.9
=1
20
y = d f (x)/dx
15
10
0
5
10
15
3
2
Abbildung 11: Fractional Derivative Calculation using modified Fourier Series. Modification still
does not work as required!
Fourier Series approximation. Extreme values are filtered
Root mean squared Error between Original signal and Fourier Series approximation
3.5
7
Original
Approximation
Before Filtration
After Filtration
1.5
0.5
2
t=1 et
2.5
q P
n
1
f (x) = x
3
2
3
2
Abbildung 12: Modified Fourier Series approximation of f (x) = x where the extreme values are
filtered based on Root Mean Squared Error threshold.

Fractional Calculus George

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fractional Calculus George

Uploaded by

Copyright:

Available Formats

Diskussion

(d) Using modified Fourier Series

et = ReconstructedF unction OriginalF unction

%-------------------------------------------------------------------------N = 100; % Number of points per period

%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Another Way %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

f_aux = inline([( InputFunction ) .*cos(w0*m*x)],x,m,w0);

f_aux = inline([( InputFunction ) .*sin(w0*m*x)],x,m,w0);

%%%%%%%%%%%%%%% Reconstruct the Signal %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%%%%%% Perform filteration of extreme values %%%%%%%%%%%%%%

Error_Method2 = sqrt(mean((Function_Reconstructed - U).^2));

ErrorVector_Method2 = zeros(1, length(t));

ErrorVector_Method2(i) = sqrt((Function_Reconstructed(i) - U(i)).^2);

Error_Method2_New = sqrt(mean((Function_Reconstructed_filtered - U).^2));

%-------------------------------------------------------------------------%%%%%%%%%%%% Recover Fourier Coefficients of Filtered Signal %%%%%%%%%%%%%%

Fourier_Series_Method2 = Fourier_Series \ Function_Reconstructed_filtered;

%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%% Reconstruct filted signals %%%%%%%%%%%%%%%%%%%%%%%%%%%%%

ErrorVector_Method_Reconstructed = zeros(1, length(t));

%-------------------------------------------------------------------------%%%%%%%%%%%%%%%%%%%%% Calculate the Partial Derivative %%%%%%%%%%%%%%%%%%%%

A0_Fourier = A0_New_filtered/(gamma(1 - q) + eps);

Example (1): f(x) = sin(x)

for alpha = -10 : 1 : 10

h = legend($\alpha = -1$, $\alpha = -0.9$, $\alpha = -0.8$, $\alpha =

-0.7$, $\alpha = -0.6$, $\alpha = -0.5$, $\alpha = -0.4$, $\alpha =

Fractional derivatives of f (x) = sin(x) of order = 1, 0.9, ..., 0.9, 1.0

Example (2): f(x) = x*(x pi)*(x 2*pi)

h = legend($\alpha = -1$, $\alpha = -0.9$, $\alpha = -0.8$, $\alpha =

%------ alpha = 0, 0th fractional derivative -----------------------------U = subs(Y,x,t);

%------ alpha = 1, 1st fractional derivative / normal derivative ---------R = diff(Y,x);

Abbildung 9: Fractional Derivative Calculation using modified Fourier Series.

f (x) = x (x pi) (x 2 pi)

Example (3): f(x) = x

h = legend($\alpha = -1$, $\alpha = -0.9$, $\alpha = -0.8$, $\alpha =

$\alpha = 0.4$, $\alpha = 0.5$, $\alpha = 0.6$, $\alpha = 0.7$,

Fractional derivatives of f (x) = x of order = 1, 0.9, ..., 0.9, 1.0

Fourier Series approximation. Extreme values are filtered

You might also like

f_aux = inline([( InputFunction ) .cos(w0m*x)],x,m,w0);

f_aux = inline([( InputFunction ) .sin(w0m*x)],x,m,w0);

Example (2): f(x) = x(x pi)(x 2*pi)