Add an overland flow solver for pipeline rupture simulations #367
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- initialize landspill module and set_landspill - python interface for specifying point source data in setrun.py - use instrinsic function new_line in generic write - add set_landspill to amr2.f90 - modify Makefile.geoclaw
- add a function to apply point sources - fix a bug in cell_id subroutine of point source instances - add function call for applying point sources to src2.f90
This type calculate the friction coefficients on the fly based on local Reynolds number and three models corresponding to laminar, smooth turbulent, and fully rough turbulent regimes. - modify the signature of get_coefficient member function - add kinematic viscosity to land_spill module - add the 4th derived Darcy-Weisbach implementation
…blems - remove misused idint function - intrinsic math function: replace single precision version with double precision version - rename three_models with three_regimes - fix the wrong signature of get_coefficient in DarcyWeisbach class
- a new class PointSourceCollection to encapsulate point sources - rename class point_source -> PointSource - rewrite I/O interface - put point-source-related source files to a sub-folder
- calculation of Reynolds number - definition of starting Re for full turbulence - DarcyWeisbachThreeModels => DarcyWeisbachThreeRegimes - semi-implicit firction term
- add flexibility for FFLAGS in the Makefile of the example - remove `make .data` and `make .exe` as `make .output` will do that - add python 2 support to download_topo.py and fix the bug of `/usr/bin/env` - set up module_setup to true after module initialization. - replace `write(*, *)` with `print *,` - check if the list of end times provided by users is sorted or not in `data.py`
- explicitly triger refinement around point sources when t = t0 - reduce the initial depth at the location of the point sources ot 0.9*dry_tolerance
Implement point sources
Capability of setting extra physical properties
Evaporation models
Landspill module and its objects do not support lat-long coordinate do far. But when these objects are null objects (e.g., zero point source, zero hydrological feature, etc), lat-long coordinate should still be allowed.
If a point source is located at the boundary grid line of a grid patch, rounding errors may cause the cell_id function not able to find correct cell indices of the point source. In this fix, we add a tolerance to the calculation.
* add settings of hidden parameters to setrun.py * modify Makefile files * revise README.md files * add a new example utah_dem_hydro_feat_opt
Optimization and landspill-specific modification
fix #18: row indices not initialized correctly
Resolve merge conflicts
remove dockerfiles for PR reviewers
remove copyright notices
Member
|
@piyueh in case you are wondering we are looking at this right now. |
Author
|
Thank you very much! |
Author
|
@mandli Hi, I'm thinking about re-working on this PR:
Is it OK if I close this PR and make a new one later? Thanks! |
Member
|
@piyueh that sounds good. Sorry we never got back to this! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Thank you all for developing this awesome software! We expanded GeoClaw to simulate the hydrocarbon overland flow released from ruptured pipelines. All extensions and modifications are located inside the folder
src/2d/shallow/landspill. None of the original files insrc/2d/shallowis modified, except the changes made in the commit042d04d. In this way, we can minimize the possibility of introducing new bugs into GeoClaw. Also, examples are inexamples/landspill.All source code of our work had been reviewed by @mandli at our fork of GeoClaw. See the following PRs in the forked repository: PR8, PR9, PR10, PR14, PR16, and PR17. We use object-oriented design to make things more modularized.
The following is an overview of our work:
1. Point sources
The releasing points of fluid from pipelines are modeled as point sources. The code is inside
src/2d/shallow/landspill/point_source. A point source is represented byPointSourceclass. And a collection of point sources is represented byPointSourceCollection. We simply add point sources to the right-hand-side of the mass equation insrc2.f90.See PR8 for more details.
2. Darcy-Weisbach friction models
We prefer Darcy-Weisbach friction models over Manning's. The source code is in
src/2d/shallow/landspill/darcy_weisbach. There are many different models to calculate the friction coefficient in Darcy-Weiscath friction. So we use an abstract class,DarcyWeisbachBase, to rule how all different models should be implemented and follow a unified interface. All models for calculating coefficients should be derived from this abstract class. Currently, we have implemented 7 different models:DarcyWeisbachNull,DarcyWeisbachConstant,DarcyWeisbachBlockConstants,DarcyWeisbachCells,DarcyWeisbachChurchill,DarcyWeisbachTwoRegimes, andDarcyWeisbachThreeRegimes.The abstract-class design requires using pointers in the main function. But pointers seem to be not welcomed in Fortran, so we use another class,
DarcyWeisbach, to encapsulate the pointers. And users of this class will not have to deal with pointers in their main Fortran code.The Darcy-Weisbach friction model replaces Manning's model in
src2.f90. See PR9 for more information.3. Intersection with On-land hydrological features
When hydrocarbon fluid contact with on-land hydrological features (e.g., rivers, lakes, ponds, etc.), we need to record the location of contact lines, contact times, and the volumes flowing into these water bodies. And after an overland flow simulation, we can use these data as the input to other solvers that simulate hydrological transport. The source code is in
src/2d/shallow/landspill/hydrologic.Users provide hydrological data through ESRI ASCII files, and one ASCII file is represented by an instance of
HydroFeatureclass. The class,HydroFeatureCollection, represents a collection of several files. Our implementation uses a concept similar to CSR sparse-matrix format to describe the locations of water bodies. Cells with zero values are land while those with values greater than zero are water bodies. Different values may refer to different types of water bodies. Now, all types of water bodies are handled in the same way, but in the future, different types of water bodies may be processed in different ways.See PR10.
4. Temperature-dependent viscosity with the Lewis-Squires viscosity correction
Many models for the Darcy-Weisbach coefficients require the viscosity of the working fluid. We use the Lewis-Squire viscosity correction to get temperature-dependent viscosity. Currently, however, due to that ambient temperature and fluid temperature do not change during a simulation, so the viscosity is actually a constant from the beginning to the end of a simulation. We determine the viscosity based on the ambient temperature provided by users.
See PR14.
5. Evaporation models
We again utilize abstract-class design to implement evaporation models. Currently, there are only two models: Fingas' log model and Fingas' square root model, which are implemented in the classes
EvapFingas1996LogandEvapFingas1996SQRT, respectively. The classEvapNullis used when evaporation is disabled, andEvapBaseis the abstract class. Again, we use a top-level classEvapModelto encapsulate pointers. We hope this kind of design can make it easier to add more models in the future.We treat the evaporation as a source (a negative source) in the right-hand side of the mass equation. And this is done in
src2.f90.See PR16 for more information.
6. Miscellaneous and minor optimization
Two slight optimizations are done in PR17. One is to ignore totally dry grid patches during time-marching. The other one is just to remove unnecessary (to oil spill applications) refining criteria in
flag2refinre2.f90so that the code can be cleaner for reading. The optimized version of the two source files is located insrc/2d/shallow/landspill/optimized.The only two modifications in the original GeoClaw source code are also in this PR (see
042d04d). The first one adds a numerical parameter toupdate.f90. This parameter controls how we update a coarse cell when the averaged fluid level of its child cells is below the coarse cell's topographical elevation. Originally in GeoClaw, when the averaged fluid level of child cells is lower than the parent cell's elevation, the solver chooses to maintain the steady-state condition rather than chooses to conserve the mass, so the parent cell will have zero fluid volume and will lose some fluid mass. But this causes some trouble for overland flows when there are canyon-like features in the topography. See the tests and the short report in this repository. After discussing this issue with @mandli, we decided to add a parameter to control the fluid volume of a coarse cell in this situation. The default value of this parameter is zero, and this will behave like the original GeoClaw. In our overland-flow applications, we set this parameter todry_tolso that a coarse cell will always have some fluid in it as long as one of its child cells has fluid. This may break the steady-state condition, but it saves us from losing significant fluid mass in our applications.The other change to GeoClaw is another numerical parameter in
flag2refine2.f90. This parameter controls the criteria to refine a cell according to the fluid depth. The default value of this parameter isdry_tol, and it behaves like the original GeoClaw. In our applications, we'll set this parameter to zero, and the finest mesh will be applied to wherever there's fluid. The area of overland flow is usually small compared to the computational domain, so using the finest mesh to cover the whole flow area is computationally acceptable.7. Examples
There are 5 examples along with their README in
examples/landspill:The first one,
inclined_planeis a validation test. We compare the result with the experimental data from Lister (1992). This test takes a very long time to run due to the very fine mesh. Be cautious.utah_demandutah_dem_hydro_featsimulate the overland flow of Maya crude somewhere close to Salt Lake City.utah_demrepresents a case where no on-land hydrological features present, whileutah_dem_hydro_feathas a pond near the rupture point.utah_dem_optandutah_dem_hydro_feat_optare the same asutah_demandutah_dem_hydro_feat, except they use optimized version ofupdate.f90andflag2refine2.f90.