Speaker
Description
As a modeller of discrete fracture network models, one usually gets tables with given orientation data gathered by mapping geologists. Seldom these data are questioned by the modeller but rather taken as definite. This may lead to spurious inferences and hence fracture models that do not accurately reflect the true condition of the rock mass.
Uncertainties are not only stemming from tool imprecision, but to a larger extent from external disturbances and human factors. By systematically identifying all relevant sources of uncertainty, that may affect the measured value, and estimate their individual contributions, it is possible to infer an aggregated probability space of the orientation for each single object. This can be expressed as a rough one-parameter dihedral angle measure, Ω, or as a non-parametric surface on a lower hemisphere stereonet, Χ.
The one-parameter representation is well suited for the analysis of large datasets, e.g. to discriminate observations with large uncertainty. In contrast, the non-parametric visualization is useful when examining individual or a few structures that e.g. are in conflict with the established conceptual understanding of the rock mass. The uncertainties may also be used for inverse modelling workflows to evaluate more accurate input parameters for fracture network models.
| Preferred duration for demo/hackathon | 20 minutes |
|---|