ode23, too slow

Question

Gloria 2022-6-3

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1733015-ode23-too-slow

评论： Steven Lord 2022-6-3

It takes a long time to solve the code. Is it because the code is reading data from a.dat file? How can I make it go faster? It slows down even more as time increases and the time step gets smaller.

2 个评论
显示无隐藏无

Walter Roberson 2022-6-3

https://www.mathworks.com/help/matlab/math/parameterizing-functions.html

Walter Roberson 2022-6-3

I also suspect that you have a stiff system and should try ode23s

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jan 2022-6-3

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1733015-ode23-too-slow#answer_977685

在 MATLAB Online 中打开

Reading the files in each iteration is a very time-consuming method. Do this once only and store the values in a persistent variable:

function [du] = mart(t,u)
persistent Fyr FyrPi Fzr FzrPi Mxr MxrPi f
if isempty(Fyr)
   Fyr   = load('fz_F_i_recon.dat');
   FyrPi = importdata('fz_Phi.txt');
   Fzr   = load('fy_F_i_recon.dat');
   FzrPi = load('fy_Phi_i_recon.dat');
   Mxr   = load('mx_F_i_recon.dat');
   MxrPi = load('mx_Phi_i_recon.dat');
   f     = load('frequency.dat');
end

Do not include all import commands in [ and ]. [] is Matlab's operator for a concatenation of arrays. With [load('file')] you concatenate the output of load() with nothing, so this is a waste of time only.

Save time by creating constants as constants: 77e9 is cheap, 77*10^9 is an expensive power operation.

There is no reason to create the same variables in a loop repeatedly:

for j=1:4
   ...
   c(1)=60;c(2)=180;c(3)=150;c(4)=100;   % Why?! Move it behind the loop
   k(1)=1800;k(2)=2400;k(3)=2000;k(4)=2000;
end

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Steven Lord 2022-6-3

Do this once only and store the values in a persistent variable:

Alternately load them outside mart.m entirely and pass them into the mart function (which is the ODE function for the system, I assume; I have not looked at all the files) as extra parameters.

请先登录，再进行评论。