Hello, I’m trying to fit transport parameters (mixing chamber dimensions and system axial dispersion) for an acetone pulse experiment I ran.
Some background - I’m evaluating small scale membrane chromatography so the pulse length is a fraction of a second. I’m having issues with the simulated results reaching the injected concentration without inputing a minimum of 1 second pulses.
Is there a way to increase the time resolution to 1/10 of a second rather than 1 second?
Also - I’m running an optimization to fit transport parameters and I would like to use chromatograms normalized by the 0th moment to account for any sensor drift or solution changes that may occur during longer run times. Is there a simple way to perform this type of analysis during the optimization? Currently I’m normalizing the experimental data by the 0th moment and creating a pulse of sufficient height to force the area under the curve to be 1.
I’d be happy to provide more information and details if the above isn’t sufficient or clear…
Cheers,
-M. Rosario Cervellere
Hi Rosario,
Regarding your first question: assuming you’re using CADET-Process, you can simply set the time_resolution attribute of the simulator to the desired interval. Please refer to the documentation for more information: Process Simulation — CADET-Process 0.10.0 documentation
I’ll think about your second question for a moment, ok?
Best regards,
Johannes
Indeed I’m using CADET-Process and adjusting the time_resolution attribute did the trick.
1 Like
Hey Rosario,
Can I ask some follow up questions regarding this part of your post?
Also - I’m running an optimization to fit transport parameters and I would like to use chromatograms normalized by the 0th moment to account for any sensor drift or solution changes that may occur during longer run times. Is there a simple way to perform this type of analysis during the optimization? Currently I’m normalizing the experimental data by the 0th moment and creating a pulse of sufficient height to force the area under the curve to be 1.
What kind of drift are you experiencing? E.g.
And which comparison metrics are you using? (SSE, NRMSE, Shape, etc.)? Depending on the drift and metric, different kinds of pre-processing and post-processing can be advisable.
1 Like
Here’s an example, I have replicates of the same flow rate and device, same feed with 2% acetone by volume, and I’m currently using NRMSE
In this case I would recommend using the Shape objective function with include_height=False. This will measure the difference in Peak-timing and Peak-shape while ignoring the height.
