Why pdetool becomes very slow after coefficient values are updated?

Question

0 votes

I am working on PDE tool to solve two coupled PDE equations. I want to run the code for each k values(k is a constant inside the a coefficient.)For example, I run pde solve for k=50 for tlist=0:0.05:100.Then if the condition met in my solution, I want to use the solution at t=100 and use it as initial condition then change the k value to 49(That is update the a coefficient) then run the code from tlist=100:0.05:120 and so on. But once the k change in the program, it becomes very slow and didn’t able to complete it even within 12 hrs. If I run the program for different k value separately then for each k value program can complete within 1hr.Since I want to run the code for 40 different k values, running the code for each k value is not possible. In the program geometry and mesh does not change and it is fixed. What is the reason for this? How can I solve it?

ti=0;step=0.05;k=50;tf=200;
u0 = @u0fun;
initcon=u0;
%Lets call for initial run
results1=GLAdeptive(k,ti,tf,step,initcon,model);
kvalue=zeros(1,11);ti1=200;tf1=205;initcon1=results1;
%Prepare for the loop
while k>30
   results2(i) = GLAdeptive(k,ti1,tf1,step,initcon1,model);% code that calculate solution for each k and time
   u1=results2(i).NodalSolution(:,1,end);
   u2=results2(i).NodalSolution(:,1,:);
   uAlongY=u1(nodesAlongY);
   uAlongY=uAlongY(ind);
   for e=1:length(uAlongY)
       partial(e)=(uAlongY(e)-mean(uAlongY))^2;
   end
   sigma=sum(partial)/length(uAlongY);
    formatSpec1 = 'sigma is %i\n';
           fprintf(formatSpec1,sigma);
   if sigma<1e-6
       initcon1=results2(i);
   i=i+1;
   ti1=ti1+5;tf1=tf1+5;
   elseif sigma>=1e-6
       clear sigma1 l
       l=1;sigma1=0;
       while sigma1<1e-6
           U=u2(:,1,l);
           uAlongY2=U(nodesAlongY);
           uAlongY2=uAlongY2(ind);
           for e1=1:length(uAlongY2)
               partial1(e1)=(uAlongY2(e1)-mean(uAlongY2))^2;
           end
           sigma1=sum(partial1)/length(uAlongY2);
           formatSpec = 'sigma1 is %i\n';
           fprintf(formatSpec,sigma1);
           l=l+1;
       end
       o=o+1;
       kvalue(o)=k;
       Le=length(results2);
       if Le>1
           Ht(o)=0.57+0.00005*l*0.05+0.00005*(results2(Le-1).SolutionTimes(101)-57);
       else
           Ht(o)=0.57+0.00005*l*0.05+0.00005*(results2(Le).SolutionTimes(101)-57);
       end
       U2=u2(:,1,l);
       uAlongY3=U2(nodesAlongY);
       uAlongY3=uAlongY3(ind);
       A=uAlongY3-mean(uAlongY3);
       numpt=length(uAlongY3);
       L=height; 
       x=sort(yline4_sort(3,:));
       k3=(2*pi/(L))*[0:numpt-1];
       signal=A; 
       utn=fft(signal); %FFT the noisy signal
       [M,I] = max(abs(utn)/max(abs(utn)));
       lamda_kappac(o)=(2*pi*k)/((2*pi)/k3(I));
       k=k-2;
       initcon1=results2(i-1);
       i=i+1;
       ti1=ti1;tf1=tf1;
       delete(model.EquationCoefficients);
   end
end