Solve equation without Symbolic Math Toolbox for Compiler

21 views (last 30 days)
I have a very long implicit equation. Actually I solve this equation using 'syms' and 'vpasolve'.
syms R % R is the only unknown variable, all others are known
g = ((1+nu)*(1-2*nu)*(1-1/(R/b0)^2))/((alpha-1)-(1-2*nu)*(1+alpha)/(R/b0)^2);
omega = -(2*(1-2*nu)*(1-nu^2))/(delta*(alpha-1))*((alpha+1/beta)-(nu/(1-nu))*(1+alpha/beta))/...
((alpha-1)-((1-2*nu)*(1+alpha)/(R/b0)^2))/(R/b0)^2;
S1 = 0; % sommatoria manuale
for n=0:1:10
if n==gam
S = k*omega^n/factorial(n)*log(R/a);
else
S = (omega^n/(factorial(n)*(n-gam))*((R/a)^(k*(n-gam))-1));
end
S1=S1+S;
end
eqn = (nn/gam)*((1-g/delta)^((beta+1)/beta)-(a0/R)^((beta+1)/beta)) == S1;
sol_R = vpasolve(eqn, R, [a0 Rpl_max]); % solution
The problem is that I want to compile this equation in a standalone application but Matlab Compiler does not support Symbolic Math Toolbox.
First I tried to convert the symbolic expression into a anonymous funcion, as follow.
syms a a0 alpha0 b0 beta0 delta0 gam k nn nu0
syms R
g = ((1+nu0)*(1-2*nu0)*(1-1/(R/b0)^2))/((alpha0-1)-(1-2*nu0)*(1+alpha0)/(R/b0)^2);
omega = -(2*(1-2*nu0)*(1-nu0^2))/(delta0*(alpha0-1))*((alpha0+1/beta0)-(nu0/(1-nu0))*(1+alpha0/beta0))/((alpha0-1)-((1-2*nu0)*(1+alpha0)/(R/b0)^2))/(R/b0)^2;
S1 = sym(0);
for n=0:1:10
%piecewise is needed to test against a value to be determined later
S = piecewise(n == gam, ...
k*omega^n/factorial(n)*log(R/a), ...
(omega^n/(factorial(n)*(n-gam))*((R/a)^(k*(n-gam))-1)) );
S1=S1+S;
end
EQN = (nn/gam)*((1-g/delta0)^((beta0+1)/beta0)-(a0/R)^((beta0+1)/beta0)) - S1;
matlabFunction(EQN, 'file', 'EQN.m', 'vars', {R, [a, a0, alpha0, b0, beta0, delta0, gam, k, nn, nu0]})
After this, I tried to solve it in different ways:
1st way) I can't use vpasolve for function_handle.
2nd way) Using fzero instead of vpasolve I can find a solution, but only if function changes sign. Therefor sometimes I get errors.
parameters = [a, a0, alpha, b0, beta, delta, gam, k, nn, nu];
obj = @(R) EQN(R, parameters);
sol_R = fzero(obj, [a0 Rpl_max]);
3rd way) Using fsolve I got same error of vpasolve.
Is there a way to reach my goal?

Answers (1)

Ameer Hamza
Ameer Hamza on 20 Apr 2020
I guess you are trying to run these lines
parameters = [a, a0, alpha, b0, beta, delta, gam, k, nn, nu];
obj = @(R) EQN(R, parameters);
sol_R = fzero(obj, [a0 Rpl_max]);
without clearing the variable EQN from your workspace. When MATLAB creates the anonymous function obj, it picks EQN from the workspace, which is a symbolic variable, and hence you get such error. Try to run the above lines after cleaning the variable EQN from workspace
clear EQN

Community Treasure Hunt

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

Start Hunting!

Translated by