Fitting two curves with shared parameters to two datasets
3 vues (au cours des 30 derniers jours)
Afficher commentaires plus anciens
I want to fit the two functions
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1538100/image.png)
to a dataset of the form
. Here
are a dimensionless distance and two angles (degrees) respectively.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1538105/image.png)
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1538110/image.png)
I generate the dataset using the above
functions with some Gaussian noise on top. The data generation seems to work fine.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1538115/image.png)
Yet, the call "lsqcurvefit(fun, x0, alpha_data_tab, beta_data_tab);" gives me an error which indicates that there are some dimension mismatch. However, I do not see it. Can anyone help me with this issue?
clc
close all
clear all
set(0,'defaulttextinterpreter','latex')
set(0,'defaultAxesTickLabelInterpreter','latex');
set(0,'defaultLegendInterpreter','latex');
set(0, 'DefaultLineLineWidth', 2);
set(0,'defaultAxesFontSize',15)
%True parameters
Af0_true = -0.017767; %Fast initial amplitude
As0_true = 0.017767; %Slow initial amplitude
lambda_f_true = -4.0571e-04; %Fast damping parameter; 1/cm
lambda_s_true = 6.4718e-05; %Slow damping parameter; 1/cm
angular_vel_f_true = 0.0228; %Angular velocity; rad/cm
angular_vel_s_true = 0.0050; %Angular velocity; rad/cm
dx = 0.01; %cm
s_true = 0:dx:1000; %cm
std = 0.1; %deg
spacing = 4;
s_data1 = 275:spacing:300;
s_data2 = 350:spacing:375;
s_data3 = 425:spacing:450;
s_data4 = 500:spacing:525;
s_data = [s_data1, s_data2, s_data3, s_data4];
[s, alpha, beta, s_data, alpha_data, beta_data] = dataMaker(Af0_true, As0_true, lambda_f_true, lambda_s_true, angular_vel_f_true, angular_vel_s_true, s_true, s_data, std);
alpha_data_tab = [s_data(:), alpha_data(:)]; % Concatenate Vectors
beta_data_tab = [s_data(:), beta_data(:)]; % Concatenate Vectors
fun1 = @(v,s_data)v(1)*exp(v(3)*s_data).*cos(v(5)*s_data) + v(2)*exp(v(4)*s_data).*cos(v(6)*s_data);
fun2 = @(v,s_data)v(1)*exp(v(3)*s_data).*sin(v(5)*s_data) + v(2)*exp(v(4)*s_data).*sin(v(6)*s_data);
fun = @(v,s_data)[fun1(v,s_data)', fun2(v,s_data)']; % Composite Function
%initial guess
percent = 0.2; %Percent deviation
x0 = [Af0_true-percent*Af0_true, As0_true-percent*As0_true, lambda_f_true-percent*lambda_f_true, lambda_s_true-percent*lambda_s_true,...
angular_vel_f_true-percent*angular_vel_f_true, angular_vel_s_true-percent*angular_vel_s_true];
fun(x0, s_data)
X = lsqcurvefit(fun, x0, alpha_data_tab, beta_data_tab);
function [s_total, alpha_total, beta_total, s_data, alpha_data, beta_data] = dataMaker(Af0_true, As0_true, lambda_f_true, lambda_s_true, angular_vel_f_true, angular_vel_s_true, s_total, s_data, std_deg)
s_total = s_total;
std_rad = std_deg*(pi/180); %rad
%True parameter vector
v0 = [Af0_true, As0_true, lambda_f_true, lambda_s_true, angular_vel_f_true, angular_vel_s_true];
alpha_total = v0(1)*exp(v0(3)*s_total).*cos(v0(5)*s_total) + v0(2)*exp(v0(4)*s_total).*cos(v0(6)*s_total);
beta_total = v0(1)*exp(v0(3)*s_total).*sin(v0(5)*s_total) + v0(2)*exp(v0(4)*s_total).*sin(v0(6)*s_total);
s_data = s_data;
alpha_data = v0(1)*exp(v0(3)*s_data).*cos(v0(5)*s_data) + v0(2)*exp(v0(4)*s_data).*cos(v0(6)*s_data);
beta_data = v0(1)*exp(v0(3)*s_data).*sin(v0(5)*s_data) + v0(2)*exp(v0(4)*s_data).*sin(v0(6)*s_data);
alpha_noise = std_rad*randn(1,length(s_data));
beta_noise = std_rad*randn(1,length(s_data));
alpha_data = alpha_data + alpha_noise;
beta_data = beta_data + beta_noise;
end
0 commentaires
Réponse acceptée
Torsten
le 13 Nov 2023
Modifié(e) : Torsten
le 13 Nov 2023
clc
close all
clear all
set(0,'defaulttextinterpreter','latex')
set(0,'defaultAxesTickLabelInterpreter','latex');
set(0,'defaultLegendInterpreter','latex');
set(0, 'DefaultLineLineWidth', 2);
set(0,'defaultAxesFontSize',15)
%True parameters
Af0_true = -0.017767; %Fast initial amplitude
As0_true = 0.017767; %Slow initial amplitude
lambda_f_true = -4.0571e-04; %Fast damping parameter; 1/cm
lambda_s_true = 6.4718e-05; %Slow damping parameter; 1/cm
angular_vel_f_true = 0.0228; %Angular velocity; rad/cm
angular_vel_s_true = 0.0050; %Angular velocity; rad/cm
dx = 0.01; %cm
s_true = 0:dx:1000; %cm
std = 0.1; %deg
spacing = 4;
s_data1 = 275:spacing:300;
s_data2 = 350:spacing:375;
s_data3 = 425:spacing:450;
s_data4 = 500:spacing:525;
s_data = [s_data1, s_data2, s_data3, s_data4];
[s, alpha, beta, s_data, alpha_data, beta_data] = dataMaker(Af0_true, As0_true, lambda_f_true, lambda_s_true, angular_vel_f_true, angular_vel_s_true, s_true, s_data, std);
%alpha_data_tab = [s_data(:), alpha_data(:)]; % Concatenate Vectors
%beta_data_tab = [s_data(:), beta_data(:)]; % Concatenate Vectors
xdata = s_data;
ydata = [alpha_data(:),beta_data(:)];
fun1 = @(v,s_data)v(1)*exp(v(3)*s_data).*cos(v(5)*s_data) + v(2)*exp(v(4)*s_data).*cos(v(6)*s_data);
fun2 = @(v,s_data)v(1)*exp(v(3)*s_data).*sin(v(5)*s_data) + v(2)*exp(v(4)*s_data).*sin(v(6)*s_data);
fun = @(v,s_data)[fun1(v,s_data)', fun2(v,s_data)']; % Composite Function
%initial guess
percent = 0.2; %Percent deviation
x0 = [Af0_true-percent*Af0_true, As0_true-percent*As0_true, lambda_f_true-percent*lambda_f_true, lambda_s_true-percent*lambda_s_true,...
angular_vel_f_true-percent*angular_vel_f_true, angular_vel_s_true-percent*angular_vel_s_true];
%X = lsqcurvefit(fun, x0, alpha_data_tab, beta_data_tab);
X = lsqcurvefit(fun, x0, xdata, ydata);
function [s_total, alpha_total, beta_total, s_data, alpha_data, beta_data] = dataMaker(Af0_true, As0_true, lambda_f_true, lambda_s_true, angular_vel_f_true, angular_vel_s_true, s_total, s_data, std_deg)
s_total = s_total;
std_rad = std_deg*(pi/180); %rad
%True parameter vector
v0 = [Af0_true, As0_true, lambda_f_true, lambda_s_true, angular_vel_f_true, angular_vel_s_true];
alpha_total = v0(1)*exp(v0(3)*s_total).*cos(v0(5)*s_total) + v0(2)*exp(v0(4)*s_total).*cos(v0(6)*s_total);
beta_total = v0(1)*exp(v0(3)*s_total).*sin(v0(5)*s_total) + v0(2)*exp(v0(4)*s_total).*sin(v0(6)*s_total);
s_data = s_data;
alpha_data = v0(1)*exp(v0(3)*s_data).*cos(v0(5)*s_data) + v0(2)*exp(v0(4)*s_data).*cos(v0(6)*s_data);
beta_data = v0(1)*exp(v0(3)*s_data).*sin(v0(5)*s_data) + v0(2)*exp(v0(4)*s_data).*sin(v0(6)*s_data);
alpha_noise = std_rad*randn(1,length(s_data));
beta_noise = std_rad*randn(1,length(s_data));
alpha_data = alpha_data + alpha_noise;
beta_data = beta_data + beta_noise;
end
0 commentaires
Plus de réponses (0)
Voir également
Catégories
En savoir plus sur Get Started with Curve Fitting Toolbox dans Help Center et File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!