OpenGeoSysUncertaintyQuantification

An uncertainty quantification toolbox for OpenGeoSys 6.

OpenGeoSys 6 (OGS6) is an open-source scientific project for the simulation of thermo-hdydro-mechanical (THM) processes in porous media. Various parameters are needed for this kind of complex coupled simulation, many of which are subject to uncertainty due to imprecise knowledge.

In simulation-aided planning of safety-related projects, the effects of those uncertainties on the results must be assessed. This toolbox is intended to provide all the necessary methods to quantify the uncertainties in a validly configured deterministic OGS6 simulation. Special care is taken to ensure reliable and accurate determination of the stochastic moments even when a large amount of data is generated for individual calculations.

The current environment for using this toolbox is individual servers. This means additional local workers are added via the distributed.addprocs method. This can be easily extended for the use in cluster environments, but further design decisions must be made first, such as requiring a Network File System (NFS), or the implementation of a mapping of executed snapshots to individual servers.

This toolbox heavily relies upon the following projects:

• DistributedSparseGrids.jl

  A library implementing an Adaptive Sparse Grid collocation method for integrating memory-heavy objects generated on distributed workers (JOSS paper).

• DistributedMonteCarlo.jl

  Analogous to the above for the Monte Carlo method. Implements Monte Carlo integration, Sobol' indices and Morris means. Morris means can be integrated alternatively by Latin Hypercube integration.

• Ogs6InputFileHandler.jl

  A simple OGS6 input file handler.

• VTUFileHandler.jl

  A VTU library for reading and writing vtu files. In addition, all mathematical operators are provided needed for stochastic postprocessing. This results in the datatype VTUFile can directly be used with the sparse grid, enabling interpolating complete OGS6 result files (JOSS paper).

• XDMFFileHandler.jl

  Analogous to the above for the XDMF file format resulting in a datatype XDMF3File.

• AltInplaceOpsInterface.jl

  A simple interface for in-place operators needed in different packages.

Since creating stochastic calculations based on deterministic solutions of discretized partial differential equations is complicated in itself, this project will focus on generating stochastic OGS6 projects. However, this project could serve as a basis for creating functionalities for generic stochastic calculations. Furthermore, the Julia projects mentioned above can be individually used to help with creating generic stochastic computations.

the package is currently under active development. If you encounter problems while using it (which is to be expected), feel free to open an issue.

import Pkg
Pkg.add("OpenGeoSysUncertaintyQuantification")

See the docs.

Contributions to or questions about this project are welcome. Feel free to create a issue or a pull request on GitHub.