| T O P I C R E V I E W |
| Stijn SWL |
Posted - 05/25/2010 : 10:19:56 AM
Origin Ver. and Service Release: OriginPro 8.0.63.988 SR6
Operating System: windows xp or windows 7
problem: fitting a complicated function with 2 independent variables and several parameters
I try to fit a both temperature (independent variable x) and pressure (independent variable y) dependent function (Troe expression) to rate coefficient data using non-linear curve fitting.
For this I use to following function with 6 parameters
independent variables: x,y
dependent variables: k
parameter names: A1,n1,E1,a,T3,T1
function (origin C):
double M = y * 0.1 / (8.31447 * x);
double kf = A1 * x^n1 * exp(-E1/(1.9858*x));
double k0 = 1.416E+59 * x^(-11.93) * exp(-83980/(1.9858*x));
double Fcent = (1-a)*exp(-x/T3) + a*exp(-x/T1);
double N = 0.75 - 1.27*log(Fcent);
double F = exp( ln(10)* log(Fcent)/(1+((log(k0*M/kf)/N)^2)) );
k = F * kf / (1+kf/(k0*M))
Compiling the function works, and I don't find any mistakes in it.
The data points for the rate coefficients (k) I like to fit, are calculated by this function itself using:
A1 = 3,02e+23
n1 = -1,88
E1 = 85710
a = 0,7646
T3 = 8,344e+9
T1 = 724,8
The dataselection is as follows:
x: A(x)
y: B(y)
k: C(y)
If I give the used values (or something close to it) as initial values for the parameters, after several iterations the data fitting stops resulting in totally different parameters (I do see however Chi-square being reduced!?).
In some occasions it says 'Fit may have converged - cannot find direction to change parameters to reduce Chi-sqr'. Or 'Fit did not converge - reason unknown'.
I checked things like making column C Z in stead of y. I started with surface instead of curve fitting, but I still didn't find a solution.
How does one solve a problem like things using Origin?
Might the problem be that the steps for the parameters Origin takes between each Chi-square calculations are too big? |
| 1 L A T E S T R E P L I E S (Newest First) |
| easwar |
Posted - 05/25/2010 : 5:49:52 PM
quote:
Originally posted by Stijn SWL
The data points for the rate coefficients (k) I like to fit, are calculated by this function itself using:
A1 = 3,02e+23
n1 = -1,88
E1 = 85710
a = 0,7646
T3 = 8,344e+9
T1 = 724,8
Hi,
What were the X and Y values you used to generate the data? Better yet, could you send your OPJ and FDF to tech support so they can try with exact same settings as you did?
Thanks,
Easwar
OriginLab |
|
|
