Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Dec 29, 2014, 3:43 a.m. EST
Exactly what do you want to do?
Exactly what do you want to do?
Ximena Paz Barrios Pizarro
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Dec 29, 2014, 12:17 p.m. EST
My model has two domains with different material. One of them has a initial concentration different to 0 and the other one has a initial concentration equal to 0. So, the solute transport is from the first domain to the second one.
I have experimental data concentration v/s time of de second domain, and I would like to determinate the diffusion coefficient of the first domain.
My model has two domains with different material. One of them has a initial concentration different to 0 and the other one has a initial concentration equal to 0. So, the solute transport is from the first domain to the second one.
I have experimental data concentration v/s time of de second domain, and I would like to determinate the diffusion coefficient of the first domain.
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Dec 30, 2014, 2:13 a.m. EST
I am not sure if Comsol has a data fitting feature, optimization yes, but I have not used it ever. Anyway, what you can do is to sweep the value of the diffusion coefficient (Parametric sweep under Solutions) and compare simulations with your data. That is the way electrochemists use the Scanning Electrochemical Microscope (SECM).
br
Lasse
I am not sure if Comsol has a data fitting feature, optimization yes, but I have not used it ever. Anyway, what you can do is to sweep the value of the diffusion coefficient (Parametric sweep under Solutions) and compare simulations with your data. That is the way electrochemists use the Scanning Electrochemical Microscope (SECM).
br
Lasse
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Mar 9, 2015, 6:32 a.m. EDT
Dear Ximena Paz Barrios Pizarro,
actually I work on quite a similar thing currently.
To get experimental data in COMSOL you can use an interpolation plot. Say you have a table with your experimental data, create the interpolation function (e.g. int1) and import your table data.
To be able to use this data to compare with a simulated curve got to >data sets and create a data set "Function1D" (other dimensions respectively). The procedure is described in the model geoelectrics (Forward Problem and Sensitivity; Model ID: 9636).
Briefly:
When creating the data set Function1D choose "int1" from the dropdown menu (i.e. your interpolation function).
Then create a 1D-plotgroup and a line graph. In the line graph choose the data set Function1D and for x-expression type in: int1(s). For x-expression type in: s.
Hope this helps!
Did you find a way to automate the parameter estimation?
Best regards,
Tobias
Dear Ximena Paz Barrios Pizarro,
actually I work on quite a similar thing currently.
To get experimental data in COMSOL you can use an interpolation plot. Say you have a table with your experimental data, create the interpolation function (e.g. int1) and import your table data.
To be able to use this data to compare with a simulated curve got to >data sets and create a data set "Function1D" (other dimensions respectively). The procedure is described in the model geoelectrics (Forward Problem and Sensitivity; Model ID: 9636).
Briefly:
When creating the data set Function1D choose "int1" from the dropdown menu (i.e. your interpolation function).
Then create a 1D-plotgroup and a line graph. In the line graph choose the data set Function1D and for x-expression type in: int1(s). For x-expression type in: s.
Hope this helps!
Did you find a way to automate the parameter estimation?
Best regards,
Tobias
Ximena Paz Barrios Pizarro
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Mar 9, 2015, 1:28 p.m. EDT
Hello Tobias K
First at all, thank you. I used a similar way to compare experimental data and the simulation model. I also used sweep parameter to compare different combination of parameter and find easily a better estimation.
An automatic way to do the optimization of this parameter that I thought is use the LiveLink for Matlab. So you can write a code that export the simulation data to Matlab and and calculate the minimum mean square and estimate another set of parameter here, then give this parameter to COMSOL, simulate here, and so on. Sadly, I couldn't activate this LiveLink, I don't know why, so I don't know if actually this idea really works.
Kind regards,
Ximena
Hello Tobias K
First at all, thank you. I used a similar way to compare experimental data and the simulation model. I also used sweep parameter to compare different combination of parameter and find easily a better estimation.
An automatic way to do the optimization of this parameter that I thought is use the LiveLink for Matlab. So you can write a code that export the simulation data to Matlab and and calculate the minimum mean square and estimate another set of parameter here, then give this parameter to COMSOL, simulate here, and so on. Sadly, I couldn't activate this LiveLink, I don't know why, so I don't know if actually this idea really works.
Kind regards,
Ximena
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Mar 10, 2015, 6:40 a.m. EDT
Dear Ximena,
thank you. That should also be possible with the Java-API. I will see if I can figure it out.
One thing needs to be added to my other post:
Beware, when you have more than one component in your model. Say your int-function is defined for component 2, then you will have to use "comp2.int1(s)" in the y-expression field.
Best,
Tobias
Dear Ximena,
thank you. That should also be possible with the Java-API. I will see if I can figure it out.
One thing needs to be added to my other post:
Beware, when you have more than one component in your model. Say your int-function is defined for component 2, then you will have to use "comp2.int1(s)" in the y-expression field.
Best,
Tobias
