can fmincon handle dynamic constraints ?

To be more clear

function A = Constraints_ineql1(Gd)
    [T_h_ex_in,T_h_out,~,~,T_c_in,~,~,~,~,Rp,~,...
    ~,alpha,Mh,kappa_c,kappa_h,M_h,M_c2,~,m_h,...
    m_CO2_O2_mix,y,y_CO2_O2,c2_stream,Mc,Phase_C2,Phase_h,P2,P,~,T2,Tcout,...
    ~,~,h_stream,Lc_k,~,~,~,~,Dc,Dh,A_xc,...
    Fluid_C2,Fluid_h,dz,m2_h,t1,t2,t3,te] = Inputs_opti; 


    Lx = Gd(1);
    Ly = Gd(2);
    Lz = Gd(3);
    Nb_element = floor(Lz / dz);
    N = Nb_element + 1;
    Nc = round((Lx + t3 - 2 * t1) / (Dc + t3)); % Number of channels per plate
    Np = round(Ly / (Dc / 2 + t2)); % Number of plates
    Plates_g = Nc * Np;  % total number of channels
    P_h = zeros(1, Nb_element + 1);
    P_c = zeros(1, Nb_element + 1);
    P_h(Nb_element+1) = P(1);
    P_c(1) = P2(2);
    T_h = zeros(1, Nb_element + 1);
    T_c = zeros(1, Nb_element + 1);
    [T_h,Twh, T_c] = T_func(Plates_g, Nb_element,T_h_out,Tcout,T_c_in,te, Rp, P, P2, T2, T_h_ex_in, Phase_h, Fluid_h, A_xc, m2_h, Dh, Lc_k, y, h_stream, y_CO2_O2, c2_stream, Phase_C2, Fluid_C2, m_CO2_O2_mix, M_h, M_c2, alpha, Mh, Mc, kappa_h, kappa_c);
    N = Nb_element + 1;
    for i = 1:Nb_element
        px1 = Properties.Pressure_drop2(y, h_stream, T_h(Nb_element-i+1), P_h(Nb_element-i+1), Phase_h, Fluid_h, Nhp, A_xc, m_h, Dh, dz, alpha, Mh, kappa_h);
        P_h(Nb_element-i+1) = P_h(Nb_element-i+2) - px1;
        if P_h(Nb_element-i+1) < 0
            break;
        end
        px = Properties.Pressure_drop2(y_CO2_O2, c2_stream, T_c(i), P_c(i), Phase_C2, Fluid_C2, Ncp, A_xc, m_CO2_O2_mix, Dh, dz, alpha, Mc, kappa_c);
        P_c(i+1) = P_c(i) - px;
    end
    A = [1.8e5 - sum(P_h(2:N)-P_h(1:N-1)), 1.8e5 - sum(P_c(1:N-1)-P_c(2:N)),0;1.8e5 - sum(P_h(2:N)-P_h(1:N-1)), 1.8e5 - sum(P_c(1:N-1)-P_c(2:N)),0; -2.5,0,1];
    %ceq = [];
options = optimoptions('fmincon', 'Display', 'iter');
[G, fval] = fmincon(@Opti_MHX, Gd, A,b, [], [], lb,ub,options); %fmincon(

 function 
 
M_HX = Opti_MHX(Gd)
   %%%expression%%%
 end
 

The A keeps on changing as the guess values Gd will change, the solver needs to chk this constraint first and then go to 'Opti_MHX' .

I just want to know how we can achieve this as we do in scipy minimize 'trust-constr' algo.

Thanks

Answer 1

[Matt J]

By "dynamic constraints" it sounds like you mean nonlinear constraints. If so, yes, that is what the nonlcon argument is for,

x = fmincon(fun,x0,A,b,Aeq,beq,lb,ub,nonlcon)

https://www.mathworks.com/help/optim/ug/fmincon.html#busog7r-nonlcon

However, the constraints are supposed to be continuous and differentiable, so your round() operations will probably invalidate that.

[Jai]

Thanks for considering. I'll have a look at that :)