Hello, I am trying to use cadetmatch to determine parameters for two breakthrough curves at different residence times. However, the results are worse fitting two curves together than fitting them individually.
If I only fit run1, the results look like this:
If I only fit run2, the results look like this:
However, if I fit these two together, the results look like this:
BTfit.txt (2.4 KB)
I attached my BT fit in the attachment. Could you please let me know what could be wrong?
I’m assuming the goal of your experiments is to determine the same parameter (e.g. column porosity) with two different experiments.
In a case where you get different results when fitting them individually compared to fitting multiple experiments simultaneously, it could indicate that there are some issues with the experiments or that your setup is maybe not suited to determine the parameters with the given set of experiments.
For example, the axial dispersion is velocity dependent, so if you have different experiments with different flow rates, it’s not possible to determine one value and individuals fits will hence inherently always be better than a combined one.
However, it’s hard to judge without further information. If you’re willing to share your config, we can have a further look.
It looks like both tied simulations are have a way too large dispersion coefficient. Have you tried a different initial value? the separately determined values should be a good start. How much do they differ between both experiments?
Try changing the following settings and see if this works for you
feature1.type = ‘Shape’
feature2.type = ‘Shape’
match_config.population = 100
match_config.stallGenerations = 40
The parameters are ka, kd, Qmax and surface diffusivity, and they should not change with different flow rates. I put my BTfit in the attachment, please let me know if you can’t see it. Thanks.BTfit.txt (2.4 KB)
Yes, I have tried different initial value and it still didn’t work. The values determined from separate experiments are pretty close. Please see below.
Parameters determined from run1 with cadetmatch:
ka/kd = 10000, Qmax = 6.2, Surface diffusivity = 1.0E-14
Parameters determined from run2 with cadetmatch:
ka/kd = 9998, Qmax = 6.9, Surface diffusivity = 1.3E-14
However, when I did cadetmatch with run1 and run2 together, the parameters are:
ka/kd = 220, Qmax = 11.1, Surface diffusivity = 5.5E-12
I have tried the settings you suggested, but it still didn’t work. Below are the fitting results.
Can you post a link to a zip file of the results folder and the two csv files you are matching against? That way I can look at the simulations and see if I can figure out what the problem is.
Due to the size limitation, I have shared the files with you in another way. Please let me know if you didn’t receive them. Thanks.
I just realized one of the problems when I tried to run the simulations. Both experiments are called main. Change the name of one of them otherwise it is trying to use a single simulation to fit both experiments which is certainly a source of weird issues.
If that does not fix the problem can you send me both run1.h5 and run2.h5 and I will look at them.
Yes, that did solve the problem! After I changed the name of one of them, the simulation was much better (please see below). Thanks Bill!