Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.soft-sys.math.mathematica > #11432
| From | "Kevin J. McCann" <kjm@KevinMcCann.com> |
|---|---|
| Newsgroups | comp.soft-sys.math.mathematica |
| Subject | Re: How to make fitting code using NDSolve? |
| Date | 2012-08-21 01:39 +0000 |
| Message-ID | <k0uoru$n7j$1@smc.vnet.net> (permalink) |
| References | <k0sroq$jn4$1@smc.vnet.net> |
| Organization | Time-Warner Telecom |
There are a couple of problems below.One is that you have a number of
subscripted variables that need to be Symbolized first. Also, your call
of model in FindFit has both the variables as inputs (ok), and the
desired output, but the latter is not defined as an output in your
definition of model.
I prefer to roll my own lsq rather than use FindFit, and have included a
notebook under separate email. Hope this helps.
Kevin
On 8/20/2012 4:16 AM, JYeom wrote:
> Hello,
> I am trying to fit some basic data to a partial differential equation- here are the data points:
>
> data = {
> {2.113, 2.915*^-7}, {2.256, 2.758*^-7}, {2.303, 2.567*^-7}, {2.493,
> 2.343*^-7}, {2.635, 2.085*^-7}, {2.825, 1.838*^-7}, {3.015,
> 1.614*^-7}, {3.205, 1.412*^-7}, {3.538, 1.199*^-7}, {3.775,
> 1.03*^-7}, {4.108, 8.173*^-8}, {4.487, 6.266*^-8}, {4.915,
> 4.695*^-8}, {5.295, 3.349*^-8}, {5.912, 2.34*^-8}, {6.672,
> 1.554*^-8}, {7.527, 7.692*^-9}, {8.524, 5.448*^-9}, {9.426,
> 4.326*^-9}, {10.23, 3.205*^-9}, {10.99,
> 2.083*^-9}, {11.99, -1.608*^-10}
> }
>
> And there are four parameters to fit in our fitting equation:
>
> model[ka_?NumberQ, kr_?NumberQ, kd_?NumberQ,
> kb_?NumberQ] := (model[ka, kr, kd, kb] =
> First[Subscript[Dye, 0] /. NDSolve[{
> Derivative[1][Subscript[Dye, 0]][
> t] ==\[VeryThinSpace]-ka Subscript[Dye, 0][t] +
> kr Subscript[Dyes, 0][t], Subscript[Dye, 0][0] == 1.22 10^-6,
> Derivative[1][Subscript[Dye, 1]][
> t] == -ka Subscript[Dye, 1][t] + kr Subscript[Dyes, 1][t],
> Subscript[Dye, 1][0] == 0,
>
> Derivative[1][Subscript[Dyes, 0]][t] ==
> ka Subscript[Dye, 0][t] - kr Subscript[Dyes, 0][t] -
> kd Subscript[CI, 0][t] Subscript[Dyes, 0][t] -
> kd /2 * Subscript[CI, 1][t] Subscript[Dyes, 1][t],
> Subscript[Dyes, 0][0] == 0,
> Derivative[1][Subscript[Dyes, 1]][t] ==
> ka Subscript[Dye, 1][t] -
> kd Subscript[CI, 1][t] Subscript[Dyes, 0][t] -
> kr Subscript[Dyes, 1][t] -
> kd Subscript[CI, 0][t] Subscript[Dyes, 1][t],
> Subscript[Dyes, 1][0] == 0,
>
> Derivative[1][Subscript[CI, 0]][
> t] == -kd Subscript[CI, 0][t] Subscript[Dyes, 0][t] -
> kd/2* Subscript[CI, 1][t] Subscript[Dyes, 1][t] -
> kb Subscript[CI, 0][t] Subscript[HDye, 0][t] -
> kb/2* Subscript[CI, 1][t] Subscript[HDye, 1][t],
> Subscript[CI, 0][0] == 3.18 10^-3,
> Derivative[1][Subscript[CI, 1]][
> t] == -kd Subscript[CI, 1][t] Subscript[Dyes, 0][t] -
> kd/2* Subscript[CI, 0][t] Subscript[Dyes, 1][t] -
> kb Subscript[CI, 1][t] Subscript[HDye, 0][t] -
> kb/2*Subscript[CI, 0][t] Subscript[HDye, 1][t],
> Subscript[CI, 1][0] == 0,
>
> Derivative[1][Subscript[HDye, 0]][t] ==
> kd Subscript[CI, 0][t] Subscript[Dyes, 0][t] +
> kd/2* Subscript[CI, 1][t] Subscript[Dyes, 1][t] -
> kb Subscript[CI, 0][t] Subscript[HDye, 0][t] -
> kb /2* Subscript[CI, 1][t] Subscript[HDye, 1][t],
> Subscript[HDye, 0][0] == 0,
> Derivative[1][Subscript[HDye, 1]][t] ==
> kd Subscript[CI, 1][t] Subscript[Dyes, 0][t] +
> kd Subscript[CI, 0][t] Subscript[Dyes, 1][t] -
> kb Subscript[CI, 1][t] Subscript[HDye, 0][t] -
> kb Subscript[CI, 0][t] Subscript[HDye, 1][t],
> Subscript[HDye, 1][0] == 0
> }, {Subscript[Dye, 0][t], Subscript[Dyes, 0][t],
> Subscript[CI, 0][t], Subscript[HDye, 0][t],
> Subscript[Dye, 1][t], Subscript[Dyes, 1][t],
> Subscript[CI, 1][t], Subscript[HDye, 1][t]}, {t, 0, 12}]])
>
> In the NDSolve, what we want to fit is Dye0[t]. So we wrote code for fitting using FindFit:
>
> fit = FindFit[data,
> model[ ka, kr, kd, kb][Subscript[Dye,
> 0]], {{ka, 10^(-4)}, {kr, 10^(-4)}, {kd, 500}, {kb, 500}},
> Subscript[Dye, 0]]
>
> However, we encountered the error message like this:
>
> FindFit::nrlnum: "The function value {-2.915*10^-7+Subscript[Dye, 0][2.113],-2.758*10^-7+Subscript[Dye, 0][2.256],-2.567*10^-7+Subscript[Dye, 0][2.303],<<16>>,-3.205*10^-9+Subscript[Dye, 0][10.23],-2.083*10^-9+Subscript[Dye,0][10.99],1.608*10^-10+Subscript[Dye, 0][11.99]} is not a list of real numbers with dimensions {22} at {ka,kr,kd,kb} = {0.0001,0.0001,500.,500.}."
>
> The desirable solution for parameter is around ka= 0.67, kr = 10^(-3), kd=1600, kb=1600.
>
> Does anyone know what I am doing wrong?
>
Back to comp.soft-sys.math.mathematica | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread
How to make fitting code using NDSolve? JYeom <cygnis@hanmail.net> - 2012-08-20 08:16 +0000 Re: How to make fitting code using NDSolve? "Kevin J. McCann" <kjm@KevinMcCann.com> - 2012-08-21 01:39 +0000 Re: How to make fitting code using NDSolve? long.he.dragon@gmail.com - 2012-10-19 06:48 +0000
csiph-web