Info

この質問は閉じられています。 編集または回答するには再度開いてください。

# Data Identification using recursive square method

1 回表示 (過去 30 日間)
George Francis 2017 年 11 月 9 日

Hello guys I have to identify data which contains input of the system and output of the system. I have to identified the parameters such as b0, b1, b2, b3, a1, a2, a3. Then I'll use transfer function and it should give me something like this:
% b0*Z^3 + b1*Z^2 + b2*Z + b3
% ------------------------------
% a0*Z^3 + a1*Z^2 + a2*Z + a3
I was able identify the data using least square method, but I'll use it as adaptive system. So I have to use recursive least square method with forgetting factor lambda. I tried to implement some functions from file exchange, but without successful result. So can somebody please help me? In the picture is example how it should looks like.
Also I attached data which I am using. Where U is input and Y is output of the system. Also my code for least square method.
Starting script
hold off
clc
clear
T=1:1:1170; %time
t=T'; %Transposition
figure(1)
plot(t,Y,'r')
hold on
plot(t,U,'g')
title('Data from system')
legend('Output','Input')
grid on
u=U; %Input
y=Y; %Output
denominator=0;
%-----------
[bd,ad] = lsf(3,3,u,y) %least square function
system = tf(bd,ad,1) %transfer function (Numerator,Denominator)
%
end
numerator = bd(1,4);
disp('Amplification of system: ')
disp(A)
disp('Numerator: ')
disp(numerator)
graph = lsim(system,u, t,'zoh');
%lsim Simulate time response of dynamic system to arbitrary inputs
figure(2)
plot(t,graph, t, y,'r') %red line in graph symbolize output of given data
title('Comparison')
legend('Identified','Given data')
grid on
Function lsf
function [B, A] = lsf(nb, na, u,y)
N = length(u);
Y = y(na+1:N);
%empty matrix for u and y
Xy=[]; Xu=[];
if(nb==1)
Xu = [u(2:N-(na-nb))];
for i=1:na
Xy = [Xy -y(na+1-i:N-i)];
end
elseif(nb==2)
for i=1:nb
Xu = [Xu u(nb+1-i:N-i-1)];
end
for i=1:na
Xy = [Xy -y(na+1-i:N-i)];
end
else
for i=1:nb
Xu = [Xu u(nb+1-i:N-i)];
end
for i=1:na
Xy = [Xy -y(na+1-i:N-i)];
end
end
X = [Xy Xu];
P = inv(X'*X)*X'*Y; %calculation of the least square method
A = [1 P(1:na)']; %denominator
B = [ 0 P(na+1:end)']; %numerator
##### 1 件のコメント-1 件の古いコメントを表示-1 件の古いコメントを非表示
John D'Errico 2017 年 11 月 18 日

@mohd albahrani - Please don't answer, when you are just trying to make a comment. Moved a non-answer into a comment:
"I'm trying to do the same but couldn't build the right code yet."

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by