ConnectO is a plug-in for the SKUA-GOCAD software to simulate channelized systems and analyze categorical realization quality.
Link to the TRAINING (sponsors only): Training Connecto_GoNURBS
Author: Guillaume Rongier ; Current contact: Pauline Collon
While channelized systems are widely recognized for their impact on fluids, simulate them is still a challenging topic. Two key aspects have to be kept in mind for channel simulation:
- Modeling channels as realistic as possible with the reproduction of their continuity as a key objective due to its impact on fluid flow.
- Integrating all the available data coming from well and/or seismic interpretations.
However those two aspects often hardly coexist in simulation methods, leading to unsatisfying results. ConnectO provides some tools to obtain results as satisfying as possible when dealing with channelized system simulation.
Channelized system simulation
ConnectO is based on a combination of Lindenmayer systems and potential fields to stochastically simulate channels conditioned to both hard and soft data. The Lindenmayer system, or L-system, is a formal grammar system widely used in plant modeling. In ConnectO, L-systems rely on their flexibility to generate straight to highly meandering channels in an object-based way. Their data conditioning is done through the attraction and/or repulsion to the data based on potential fields.
(Unconditional (left) and conditional (right) channel realizarions in a canyon. The channel and the canyon have been generated using a L-system and NURBS from the GoNurbs plug-in)
Related publications: A geostatistical approach to the simulation of stacked channels, Stochastic simulation of channelized sedimentary bodies using a constrained L-system
Realization quality analysis
When several simulation methods are available, the difficulty is to sort the various realizations and only retain the method that gives the best possible realizations depending on the studied case. ConnectO provides a set of functions to analyze the quality of categorical realizations. Those functions are based on quality indicators computed from connected components and their skeletons. Those indicators are used to compute dissimilarity matrices for an easy comparison of the related realizations.
(400 realizations generated with different methods compared with a training image using connected component-based indicators, dissimilarity computation and multidimensional scaling)
Related publication: Comparing connected structures in ensemble of random fields