Author Topic: Persistent, assignable initial conditions, et al.  (Read 1145 times)

scubohuntr

  • Posts: 9
    • View Profile
Persistent, assignable initial conditions, et al.
« on: February 22, 2013, 12:44:31 am »
I do a lot of minor tweaking and repeated runs of models during the calibration process (necessary because PEST is practically unuseable), and I can't count the number of times I have been burned because my initial conditions have gotten altered due to a partial run or an unintended save.

Would it be so difficult to set initial conditions as a model variable? I run my model in steady state and use the output as initial conditions for the transient model. I tried setting all my pumps to start at t=0.001 and running steady state at t=0 prior to the transient run, but that unassigns the dozens of time series in my transient model. When I go back to transient, I have to reassign all the wells. There is the possibility of importing, but really, does that make any kind of sense? I have gone to a great deal of effort to define my starting conditions, and I would just like to be sure they are staying where I put them. Even if there were a menu selection to specify a steady state .dac file to use for initial conditions, it would help.

Number two. How about some more tools for editing observation points? Maybe be able to assign a point to a specific node, rather than having to delete and redefine an observation point to move it up a slice. When I have hundreds of observation points, I like to keep pumping wells together with their observation wells. Makes exporting data much easier, particularly since the default display is the arbitrary numbering of the software, not the label I assigned to the point. If I move a point, it goes to the end of the list. Now, instead of remembering to export points 64 and 65, I need to export 64 and 297.

Probably fantasy, but would it be possible to define sets of commonly used conditions (e.g. Aquifer X) and use them to assign and adjust parameters? If I defined aquifer x as 1.0E-05 [10-4 m/s] (don't even get me started on appropriate units!!!) for Kxx and Kyy, 1.0E-06 [10-4 m/s] for Kzz, and 1E-06 1/m for Ss, wouldn't it be great to have some way to assign additional elements to the same parameters and not worry about missing a recent change? If I decide that my Ss is too low, I could just change the aquifer x definition and all the elements assigned to that definition would be updated. I have a model that has to represent subcrops in a major river valley, so some of my lithologies are represented on sixteen slices. Nudging the parameters of those units is not at all a trivial exercise, particularly since my stored selections seem to go away with no warning.

Just wishing.

Pete Sinton

  • Posts: 154
    • View Profile
Re: Persistent, assignable initial conditions, et al.
« Reply #1 on: March 04, 2013, 09:34:50 pm »
I think your Number 3 is what I would call the zone approach.  Define zones as selections of elements (or nodes) and keep a database (simple lookup table) of unique zone numbers or names, so that the zone number (or name) refers to the selection of elements/nodes.  Then can just edit the material (or node) properties in the database lookup table rather than having to activate saved selections one at a time to make changes.  A very simple and effective approach used by many.

mwtoews

  • Posts: 21
    • View Profile
Re: Persistent, assignable initial conditions, et al.
« Reply #2 on: December 01, 2013, 10:03:08 pm »
I'm pretty keen on feature one described above for assignable initial conditions. Essentially, it would be an option to specify a nodal dataset of initial head, concentration or heat for each node. This dataset would be written to all nodes before running any simulation. It is a very common feature for other modelling software. The current behaviour is to use the previous simulation state as initial conditions, which does not have any rational basis.

Currently, it appears that the best method is, e.g.:
  • Export the nodal conditions for a timestep that you like, e.g. right-click "Hydraulic head" > Export Data > All Nodes ..., and choose "ASCII Table (*.dat)"
  • From Maps, right-click and "Add Map(s)...", choose the same file
  • Right-click the dataset filename, and choose "Link to Parameter..."
  • From the available parameters from the file on the left side, find the one for the initial conditions, e.g. FINT for head values, and double-click the appropriate process variable on the right side
  • On the lower part of the dialog, set "Node/Slice Selection" to "Field containing node number", and set "Field containing node number" to "NODE"

Then, before each simulation and for each process variable (hydraulic head, mass concentration, heat):
  • In a 3D view, select all nodes for the process variable, e.g. Hydraulic head
  • From the data assignment dialog at the top, toggle a mapped assignment, choose the appropriate map and click "Put value"
« Last Edit: December 01, 2013, 10:07:32 pm by mwtoews »