Integro-Differential Equation: Analytical Approach?

  • Context: Graduate 
  • Thread starter Thread starter saltydog
  • Start date Start date
  • Tags Tags
    Interesting
Click For Summary
SUMMARY

The discussion focuses on the analytical and numerical approaches to solving integro-differential equations (IDEs), specifically a modified Runge-Kutta method. The equation presented is df/dx = -af/x^2 (∫(u^2 f(u) du) + M) with initial conditions. The user successfully reformulates the IDE into an initial value problem (IVP) and implements a numerical solution that shows close agreement between results. The conversation highlights the challenges of differentiating integrals in certain IDEs and suggests that direct numerical methods may be beneficial in those cases.

PREREQUISITES
  • Understanding of integro-differential equations (IDEs)
  • Familiarity with numerical methods, particularly the modified Runge-Kutta method
  • Knowledge of initial value problems (IVPs)
  • Basic calculus, including integration and differentiation
NEXT STEPS
  • Study the modified Runge-Kutta method for solving integro-differential equations
  • Explore analytical techniques for solving specific types of IDEs
  • Research numerical stability and error analysis in numerical methods
  • Examine case studies or articles on integro-differential equations and their applications
USEFUL FOR

Mathematicians, physicists, and engineers interested in solving integro-differential equations, as well as students and researchers looking to enhance their numerical analysis skills.

saltydog
Science Advisor
Homework Helper
Messages
1,590
Reaction score
3
I'm looking at an integro-differential equation prompted by a HW question. This is a simplified version of it:

\frac{df}{dx}=-\frac{af}{x^2}\left(\int_{x_0}^x u^2f(u)du+M\right);\quad f(x_0)=1

with:

x_0>0

and a and M positive constants.

Does anyone know if it can be approached analytically and how? I'm currently working on a modified Runge-Kutta method to solve it numerically . Is that the best approach for such equations?
 
Physics news on Phys.org
Alright, so I can rearrange to:

-\frac{x^2f^{'}}{af}=\int_{x_0}^x u^2f(u)du-M

and differentiate both side to get:

-\frac{af(x^2f^{''}+2xf^{'})-ax^2(f^{'})^2}{(af)^2}=x^2f

Simplifying I arrive at the following IVP:

xff^{''}+2ff^{'}-x(f^{'})^2+axf^3=0;\quad f(x_0)=y_0;\quad f^{'}(x_0)=-\frac{af(x_0)}{x_0^2}M
 
Last edited:
Just want to close on this one:

I modified Runge-Kutta to integrate the developing integral at each time step and add it to the results of the calculations in an effort to solve directly, the IDE numerically for the equation:

\frac{df}{dx}=-\frac{0.5f}{x^2}\left(\int_1^x u^2f(u)du-1\right)

That is, calculate the 4 slopes that are Runge-Kutta, and for each slope calculated, integrate the solution so far determined from x=1 to x=current time step. Then factor this into the calculations for the slopes.

I next calculated numerically, the equivalent IVP:

xff^{''}+2ff^{'}-x(f^{'})^2+0.5xf^3=0;\quad f(1)=1;\quad f^{'}(1)=-0.5

and ran them both to x=20. The first plot is a superposition of both results. The agreement is close. The second plot is the error of the back substitution of the first calculation. Note how the error is largest at the beginning of the calculation due to what I suspect is insufficient knowledge of the integral during that time interval.

I'm thinking there are some IDEs where you can't simply differentiate-out the integral so easily as above and therefore, the direct numerical approach may have some utility in those cases. For example, one like:

f+\frac{df}{du}=\int_0^x f(\frac{u}{q})dq

right?
 

Attachments

  • IDE (composite).JPG
    IDE (composite).JPG
    4.2 KB · Views: 449
  • IDE (discerpancy).JPG
    IDE (discerpancy).JPG
    4.6 KB · Views: 481
I would like to solve integro-differential equation with Runge-Kutta, but I don't know, how can I begin it. Do you have some articles or examples, where I can understand, how I solve an integro-differential equation. I don't know what can I do with integral. Thank you for your reply.
 

Similar threads

  • · Replies 65 ·
3
Replies
65
Views
8K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 6 ·
Replies
6
Views
5K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 10 ·
Replies
10
Views
1K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 5 ·
Replies
5
Views
3K