Unable to solve nonlinear equation using fsolve as the message shows No solution found
3 views (last 30 days)
Show older comments
function F = FeMnC660newuipffc(x)
F(1) =-9.135789053E+00-log(x(1))+166.2509975*x(1)+5.84074229*x(2)+(-166.2509975*x(1)^2)/2-5.84074229*x(1)*x(2)+(-19.52527074*x(2)^2)/2;
F(2) =-3.51500942E+00-log(x(2)) +5.84074229*x(1)+19.52527074*x(2)+(-166.2509975*x(1)^2)/2-5.84074229*x(1)*x(2)+(-19.52527074*x(2)^2)/2;
end
unable to solve this equation using fsolve as the message shows
No solution found.
fsolve stopped because the last step was ineffective. However, the vector of function
values is not near zero, as measured by the value of the function tolerance.
Kindly help me out
0 Comments
Accepted Answer
John D'Errico
on 6 Feb 2023
Edited: John D'Errico
on 6 Feb 2023
I would not be at all surprised if it was poor starting values that might cause the problem.
syms x y
F1 = -9.135789053E+00-log(x)+166.2509975*x+5.84074229*y+(-166.2509975*x^2)/2-5.84074229*x*y+(-19.52527074*y^2)/2;
F2 =-3.51500942E+00-log(y) +5.84074229*x+19.52527074*y+(-166.2509975*x^2)/2-5.84074229*x*y+(-19.52527074*y^2)/2;
fimplicit(F1,[0,5])
hold on
fimplicit(F2,[0,5])
xlabel x
ylabel y
It is the intersection of the red and blue curves where you will find a solution. There appear to be three such intersections, and a third non-solution happens at x=y=0. but that point is a singularity, not a true solution. The other solutions at x==0 and y==0 are also problematic.
But fsolve should have no real problem in finding the solution as found by @Matt J, as long as you give it reasonable starting values. A problem may be that if you are not careful, is if fsolve ever tries to go outside of the legal search space where x>0 and y>0, then fsolve will fail.
See that even vpasolve fails to find a happy solution, if you allow it to choose its own starting values.
[X,Y] = vpasolve(F1,F2,[x,y])
However, other starting values seem to get to a happy place.
[X,Y] = vpasolve(F1,F2,[x,y],[1 1])
So it was very likely a poor choice of starting values that caused the fsolve failure for @Vikash Sahu.
Can the problem be modified to avoid the issue completely? Well, yes. Replace each of x and y with the squares of two variables.
syms xx yy
G1 = subs(F1,[x,y],[xx^2, yy^2]);
G2 = subs(F2,[x,y],[xx^2, yy^2]);
[XX,YY] = vpasolve(G1,G2,[xx,yy])
X = XX^2
V = YY^2
Now fsolve should be more robust too, even with random starting values, though it still might get trapped in one of the "solutions" at x==0 or y==0.
0 Comments
More Answers (2)
Alan Stevens
on 6 Feb 2023
Edited: Alan Stevens
on 6 Feb 2023
fminsearch makes a reasonable attempt:
F1 = @(a,b) -9.135789053E+00-log(a)+166.2509975*a+5.84074229*b+...
-166.2509975*a^2/2-5.84074229*a*b-19.52527074*b^2/2;
F2 = @(a,b) -3.51500942E+00-log(b)+5.84074229*a+19.52527074*b+...
-166.2509975*a^2/2-5.84074229*a*b-19.52527074*b^2/2;
F = @(x) norm(F1(x(1),x(2))) + norm(F2(x(1),x(2)));
opt = optimset('TolFun', 1E-15);
x0 = [1, 1];
[x, fval] = fminsearch(F, x0, opt);
disp(x)
disp(fval)
Matt J
on 6 Feb 2023
Edited: Matt J
on 6 Feb 2023
You haven't shown how the optimization was executed. fsolve appears to work fine below:
opt = optimoptions('fsolve', 'FunctionTol',1E-15,'OptimalityTol',1e-15,'StepTol',1e-15);
x0 = 100*rand(1,2);
[x, fval] = fsolve(@FeMnC660newuipffc, x0, opt)
function F = FeMnC660newuipffc(x)
F(1) =-9.135789053E+00-log(x(1))+166.2509975*x(1)+5.84074229*x(2)+(-166.2509975*x(1)^2)/2-5.84074229*x(1)*x(2)+(-19.52527074*x(2)^2)/2;
F(2) =-3.51500942E+00-log(x(2)) +5.84074229*x(1)+19.52527074*x(2)+(-166.2509975*x(1)^2)/2-5.84074229*x(1)*x(2)+(-19.52527074*x(2)^2)/2;
end
See Also
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!