Im trying to run fsolve inside a for loop but it seem that my second iteration inside the for loop is not able to read the initial trial point

Code below:

clear;
clc;
%%
Disp = [1 10 4 2 3;
    2 8 3 2 2;
    3 5 4 1.5 3];

options = optimset('TolFun',1e-12,'MaxFunEvals',1e9,'Maxiter',1e9);
options.StepTolerance =1e-12;
%%
N = size(Disp,1);

O = [0,0];
tau = 0.3;
Bdispy = 0;
Cdispy = 0;
GG = zeros(2,N);
exitflags = zeros(1,N);

for i = 1:N
    
    Peakdispx(i) = Disp(i,2);
    Peakdispy(i) = Disp(i,3);
    Adispy(i) = Peakdispy(i);
    mslope(i) = (Disp(i,5)-O(2))/(Disp(i,4)-O(1));
    Cdispx(i) = Peakdispx(i)-((Peakdispy(i)-Cdispy)/(mslope(i)));
    Bdispx(i) = Cdispx(i)*-1;
    Adispx(i) = (Adispy(i)-Bdispy)/(mslope(i))+Bdispx(i);
    slopeACdisp(i) = ((Adispy(i)-Cdispy)/(Adispx(i)-Cdispx(i)));
    Afull(i) = (0.5*abs(Adispx(i)*Bdispy+Bdispx(i)*Cdispy+Cdispx(i)*Peakdispy(i)+Peakdispx(i)*Adispy(i)-Adispy(i)*Bdispx(i)-Bdispy*Cdispx(i)-Cdispy*Peakdispx(i)-Peakdispy(i)*Adispx(i)));
    
    
    F = @(GG)[slopeACdisp(i)-((GG(2)-Cdispy)/(GG(1)-Cdispx(i)));
        -tau+(0.5*abs((GG(1)*Bdispy+Bdispx(i)*Cdispy+Cdispx(i)*Peakdispy(i)+Peakdispx(i)*GG(2)-GG(2)*Bdispx(i)-Bdispy*Cdispx(i)-Cdispy*Peakdispx(i)-Peakdispy(i)*GG(1))))/Afull]
    gg0 = [tau*Disp(i,3);(1-tau)*Disp(i,2)];
    GG = fsolve(F,gg0);
    
    %% pinch in x and y
    disppinchy(i) = GG(2)/Disp(i,3);
    U(i) = ((GG(2)-Cdispy)/mslope(i))+Cdispx(i);
    L(i) = ((GG(2)-Bdispy)/mslope(i))+Bdispx(i);
    disppinchx(i) = (GG(2)-L(i))/(U(i)-L(i));
    
    
    end
for i = 1:N
    Gdispx(i) = GG(1);
    Gdispy(i) = GG(2);
    Apin(i) =0.5*abs(Gdispx(i)*Bdispy+Bdispx(i)*Cdispy+Cdispx(i)*Peakdispy(i)+Peakdispx(i)*Gdispy(i)-Gdispy(i)*Bdispx(i)-Bdispy*Cdispx(i)-Cdispy*Peakdispx(i)-Peakdispy(i)*Gdispx(i));
    ratio(i) = Apin(i)/Afull(i)
end

Answer 1

[Matt J]

Should the division be by Afull(i)?

    F = @(GG)[slopeACdisp(i)-((GG(2)-Cdispy)/(GG(1)-Cdispx(i)));
        -tau+(0.5*abs((GG(1)*Bdispy+Bdispx(i)*Cdispy+Cdispx(i)*Peakdispy(i)+Peakdispx(i)*GG(2)-GG(2)*Bdispx(i)-Bdispy*Cdispx(i)-Cdispy*Peakdispx(i)-Peakdispy(i)*GG(1))))/Afull(i)]

[John Harvie delos Reyes]

Oh, I completely missed this one.

Follow up question, how can i increase accuracy?

I often arrive at no solution found if i use my actual data for the Disp array

[Matt J]

Since it's a 2D function, you can do a grid search for a more accurate initial guess.