Hi Erik,
I once supervised a bachelor thesis that tried to do symbolic
differentiation. It was very limited, as many laws in groundwater flow
are not differentiable, either directly or by regulation near extrem
values like S close to 0 or 1. Working around with splines or similar
methods sounds tedious.
Using numerical differentiation as suggested by Timo seems to be more
promising.
Maybe the dissertation of Karin Maria Erbertseder is helpful, she did a
sensitivity analysis in section 8.6.2 for her coupled DuMuX code.
https://elib.uni-stuttgart.de/items/7e9b13f2-2a1c-4b53-98bb-bf19102eb950
Bye
Christoph
Am 26.08.25 um 09:37 schrieb Erik Kopp:
Dear Dumux Team,
I am currently beginning with Dumux and trying to simulate processes in
the soil.
Currently I am focusing on microbial processes (i.e. only ODEs) but I
want to keep using dumux so that I can keep using the results in the
full-scale simulation of the soil processes.
In order to fit the parameters, I want to compute local sensitivities
for the parameters, so derivatives of the form \frac{\partial
solution(time t)}{\partial parameter_i}.
When looking through your old mails I found one from mai of 2020 (it has
a different topic):
From dmitry.pavlov at outlook.com Fri May 15 18:25:33 2020
From: dmitry.pavlov at outlook.com (Dmitry Pavlov)
Date: Fri, 15 May 2020 19:25:33 +0300
Subject: [DuMuX] Analytic Jacobian in the case of variable viscosity
Message-ID:
<he1p191mb025295a4da86fc5f5043d245ff...@he1p191mb0252.eurp191.prod.outlook.com>
In its answer on the same day Timo Koch wrote
“(A[utomatic ]D[ifferentiation] would be an option but currently not
implemented in Dumux)”
Has automatic differentiation been implemented since?
And do you know of an automatic way to calculate these local
sensitivities? Or has someone else tried it before in another
application using dumux?
Kind Regards,
Erik
--
Most customers will not accept source code with compile errors in it.
Dan Saks, CppCon 2016 (https://youtu.be/D7Sd8A6_fYU)
_______________________________________________
DuMux mailing list
[email protected]
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
