Issue about simulation time

A discussion forum for users of the new HYDRUS 2D/3D. HYDRUS is a software package for simulating water, heat and solute movement in two- and three-dimensional variably saturated media. Happy Posting!
Post Reply
hydrokai
Posts: 13
Joined: Fri Feb 14, 2020 5:21 pm
Location: USA

Issue about simulation time

Post by hydrokai » Fri Aug 21, 2020 6:59 am

Hi all,

I am trying to use the inverse solution to optimize the hydrological parameters in van Genuchten model (1980) and HCF model (Mualem, 1976). After setting up the model and running it, the command window shows quite weird values under "Time" tab (please see below), which shows the time is jumping between 0 to 0.333333E-5. Can anyone advise on this issue? Thank you.

Best,
H.K.
Attachments
Capture.PNG
Capture.PNG (35.58 KiB) Viewed 225 times

Jirka
Posts: 5031
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Issue about simulation time

Post by Jirka » Fri Aug 21, 2020 3:42 pm

From the information that you provided, I assume that you have in the objective function one data point for time zero. That obviously should not happen. Information about time zero should be used in the initial conditions (and not in the objective/calibration function). The data points in the objective functions must be for times larger than t_init+dt_init (initial time + initial timestep). J.

hydrokai
Posts: 13
Joined: Fri Feb 14, 2020 5:21 pm
Location: USA

Re: Issue about simulation time

Post by hydrokai » Fri Aug 21, 2020 6:08 pm

Thank you Jirka for the help.

In the objective function, I have multiple data point with the time of the first one > 0.
data for inverse solution.PNG
data for inverse solution.PNG (25.84 KiB) Viewed 212 times
As suggested, I changed the initial time+initial time step so that the time of the first data point in objective function is times larger than the initial time+initial time step.
time input.PNG
time input.PNG (15.91 KiB) Viewed 212 times
When I run it, the software still gives a weird values under "time" tab, still jumping between 0 to 0.333333E-5. Please see below for the details.
command window.PNG
command window.PNG (31.28 KiB) Viewed 212 times
Thank you again for your help.

Best,
H.K.

Jirka
Posts: 5031
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Issue about simulation time

Post by Jirka » Fri Aug 21, 2020 6:18 pm

There must be some reason why the code tries to go back to zero after the first step. You must request it by some input (time in the objective function, print time, etc). Otherwise, the code would not do that. J.

hydrokai
Posts: 13
Joined: Fri Feb 14, 2020 5:21 pm
Location: USA

Re: Issue about simulation time

Post by hydrokai » Fri Aug 21, 2020 8:25 pm

Thank you very much for your help, Jirka! Will check the objective function, output etc.

Best,
H.K.

DandD
Posts: 18
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Issue about simulation time

Post by DandD » Mon Sep 07, 2020 9:10 am

Hi H.K.,
I am currently having a similar problem and one of my colleagues also had this problem this month. Could you resolve it by now? A couple things we observed:
1) The model will run through the whole simulation time one time -> when it starts the second iteration with time 0 again it starts oszillating around time 0
2) My collegaue could resolve the isse by reducing the maximum time step to a value closer to the inital time step (which lead to quite long simulation times but the model was running) & a little "trick" which somehow fixed it: say the final simulation time would be 60 min and you have time variable BCs -> add another value at the end which is similar to the BC at the start of the simulation (e.g. 61 min)

So basically what I did was defining the intial state via BCs after my intended final simulation time so that the end more or less equals the start -> this did the trick and the inverse solution is running without oszillating around 0. For the results I will just ignore this "extra part" in the end. My model is currently running so I do not know for sure if the whole issue is resolved...I still wanted to share this information at this state as it could be a bit different with your model.
EDIT: It run through but I still have another problem/error -> see thread: viewtopic.php?f=3&t=3711

A question to Jirka/Mirek: what time step is used for the start of each new iteration in the inverse solution? The originally defined initial one?

Best regards
Frieder

Jirka
Posts: 5031
Joined: Sat Mar 16, 2002 3:47 pm
Location: USA
Location: Riverside, CA

Re: Issue about simulation time

Post by Jirka » Mon Sep 07, 2020 10:08 am

You need to give the model flexibility to hit “print times” (when the output is requested), “BC Times” (when BCs are specified), and “inverse times” (when data for the inverse objective function are defined). These times have to be at least dt_min appart and none of them can be smaller that t_init+dt_init. The max allowed time step (i.e., dt_max) has nothing to do with any of this.

The initial time step (dt_init) is used at the beginning of each run, as well as when boundary conditions change (by more than some predefined changes in BCs).

J.

DandD
Posts: 18
Joined: Wed Jan 15, 2020 12:11 pm
Location: Germany

Re: Issue about simulation time

Post by DandD » Mon Sep 07, 2020 10:18 am

Thanks Jirka for the explanation - I understand that this should not make a difference but for some reason it seemed to be working. I only added this "extra" time at the end and did not change anything else in the time information (except for final time obviously), observed data or anything else. Then this 'oszillating' around 0 did not occur. I am still having this other problem (see my newest topic).
Best, Frieder

Post Reply