Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Time delay equations

  1. Sep 11, 2005 #1
    i need to solve an equation of the form (5) from this webpage http://chaos.phy.ohiou.edu/~thomas/chaos/mackey-glass.html [Broken]
    first it asks to "derive the corresponding eigenvalue equation"- i presume they mean to derive (6)? if so, i dont know how they get it.
    and then they ask for something similar to deriving (7), i have no idea how to proceed... thanks.
    Last edited by a moderator: May 2, 2017
  2. jcsd
  3. Sep 13, 2005 #2


    User Avatar
    Science Advisor
    Homework Helper

    Jesus dude, I gotta' get that book! One way to proceed of course is . . . to check out the book from a library. Deriving the eigenvalue equation is just substituting the assumed solution:

    [tex]y(t)=e^{\lambda t}[/tex]

    into the "linearized equation:

    [tex]y^{'}=\alpha y+\beta y_{\tau}[/tex]

    remembering [itex]y_{\tau}=e^{\lambda(t-\tau)}[/tex]

    Deriving (7) I assume means to solve for lambda in:

    [tex]\lambda=\alpha+\beta e^{-\lambda \tau}[/tex]

    assuming lambda is complex and determining under what conditions the real part is less than zero (not sure though, just my assumption). Really, this would take me days to fully study, a week maybe. But very interesting. Thanks. :smile:
    Last edited by a moderator: May 2, 2017
  4. Sep 14, 2005 #3


    User Avatar
    Science Advisor
    Homework Helper

    You know we can make progress with this. First put it into standard DDE form:


    That gives the rate of change of the density of white blood cells circulating in the blood as a function of the current density as well as the density at a previous time. This is doable. First thing to note that with DDEs, rather than an initial point given as the initial condition, an initial function has to be given in the interval:


    So we'll call that initial condition [itex]f_1(t)[/itex]

    Now, in the interval [itex](0,\tau)[/itex], we have a regular ODE:


    Which we can integrate from 0 to [itex]\tau[/itex]. We'll call that function [itex]f_2(t)[/itex]. Now, plug that into the DDE:


    and integrate from [itex]\tau[/itex] to [itex]2\tau[/itex]. See what's happening? Keep doing that. It get's messy. And how can this DDE help model the onset of lukemia? Think I'll spend some time on it.
    Last edited: Sep 14, 2005
  5. Sep 14, 2005 #4


    User Avatar
    Science Advisor
    Homework Helper

    Here's an example.

    [tex]\frac{dy}{dt}=\frac{0.2 y(t-14)}{1+y(t-14)^{10}}-0.1y(t);\quad y(<0)=0.5[/tex]

    As I stated earlier, we'll integrate in intervals of the delay using the previous solution as the delay functions in the ODE. Here is the first interval in Mathematica:


    [tex]\text{sol1=NDSolve}[{y^{'}==\frac{0.2 f_0[t-14]}{1+f_0[t-14]^{10}}-0.1y(t),y[0]=f_0[0]},y,\{t,0,14\}][/tex]


    Here's the second one:

    [tex]\text{sol2=NDSolve}[{y^{'}==\frac{0.2 f_1[t-14]}{1+f_1[t-14]^{10}}-0.1y(t),y[14]=f_1[14]},y,\{t,14,28\}][/tex]


    Note how I substituted [itex]f_1(t-14)[/itex] into the ODE to represent the delay for the second interval. Here's the third interval:

    [tex]\text{sol3=NDSolve}[{y^{'}==\frac{0.2 f_2[t-14]}{1+f_2[t-14]^{10}}-0.1y(t),y[28]=f_2[28]},y,\{t,28,42\}][/tex]


    And so on for each interval. A plot of the first three intervals is attached. Is there another way to do this? Should I go over to the ODE forum and ask as this isn't my homework?

    Attached Files:

  6. Sep 26, 2005 #5
    thank you!
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook