| T O P I C R E V I E W |
| AndreDourado |
Posted - 05/04/2015 : 12:44:36 PM
Hi,
I wrote, on Origin the function y=A*exp(-B*x)+C*x*exp(-B*x).
It worked twice, but with other data the report error said that it could not converge and that the reason was unknown. Once it said that the parameter were not independent and that I should try to fix one of them, what is not possible to me. What can I do?
I do not know if there is any problem with the code, or something like this.. if there is any forbiden calculation with this function...
Thank you!
André |
| 4 L A T E S T R E P L I E S (Newest First) |
| AndreDourado |
Posted - 05/08/2015 : 08:46:42 AM
The values of A can be high as 1E4 (my initial guess), or more. The value of B is something as a value normalized by gas constant and the temperature, but this value is dependent of the system in study. So I have tried to guess an initial value as 0.01.
C, as I said before, is the sum of two values, being one of them A.
Origin reported to me:
Fit may have converged - cannot find direction to change parameters to reduce Chi-sqr.
And I could not get any fit...
Sorry if I am asking silly questions, I have no knowlege on programming.
Thank you!! |
| AndreDourado |
Posted - 05/08/2015 : 08:26:52 AM
Thank you very much!
This equation came from a chemical kinetic problem. C is actually A+(other rate constant). I do not know if it is possible to link the paramther A with C on that way. One I've tried to write something like this and origin did not run because it was to manny paramthers. But I think that was on Origin 6.0, I do not know if the newer versions can do this.
I am going to try it and that I will post what I am going to get.
Thank you for the help! |
| Hideo Fujii |
Posted - 05/04/2015 : 4:23:58 PM
Hi,
I think Zheng is right.
See the sample below, which shows three results with the simulated data from A=0.8, B=2, C=1 with random noise. Once A gets a large enough value, the curve appears almost exponential, and C contributes very insensitive way. You notice that the Dependency values in the last test, where all parameters are free are very high. This should be the reason why Origin output the warning message to suggest to fix one parameter. In such case, how can you be convinced the validity of C? and if you have a boundary condition of C, it would be more plausible, I think.
Hope this help you think further.
--Hideo Fujii
OriginLab |
| lkb0221 |
Posted - 05/04/2015 : 3:43:01 PM
Hi,
It looks like you need to add some initialization code or parameter boundaries for the fitting. For some of your data, the fitting process just cannot find a good start.
Zheng
OriginLab |
|
|
