This paper introduces a new methodology for blockwise structured curvilinear grid generation for geological modeling. This approach is semi-automatic and based on triangulated surface modeling. The grid created accurately respects the geometry of the geological structures studied: it fits exactly through the layer interfaces (horizons); free-form faults, pinchouts and some unconformities can be taken into account and are faithfully represented within the grid. The underlying techniques are new in the geoscience fields. They have been initially developed for a geomechanical application but they can be useful in other geological applications. They are being used for stratigraphic grid generation. This approach is illustrated in the paper through the use of, an actual case study for a geomechanical application, and synthetic objects for stratigraphic modelling and reservoir flow simulation.
3-D grid generation plays a crucial role for numerical simulations in various geological applications. Many investigations have been made in this topic, mostly in the reservoir engineering domain because of its great interest to the oil industry. These have led to numerous 3-D finite-element or finite-difference gridding techniques and software packages depending on the application and the constraints of the simulation concerned. However, most of the proposed gridding techniques and hence simulations are applied solely to simple geological models. They rarely account for geological irregularities such as faults and pinchouts. When they do, these features are often greatly simplified. So they are not accurately represented. For instance, faults are often vertical and at most are skew but planar in the grid representation. Free-form faults are not represented. One of the main reasons is that efficient and accurate 3-D gridding is closely related to and strongly depends on efficient and accurate 3-D surface modeling. Until now, little attention has been paid to the surface modeling part in most of the existing geoscience gridding software packages. Thus it is not always possible to easily fit a surface onto measured data and to perform topological operations on surfaces. This is a common problem in almost all geoscience fields, and a great deal of effort is being made nowadays to solve this by improving surface modeling.
In our contribution, we propose a new method for blockwise structured curvilinear grid generation. This approach is semiautomatic and based on Gocad triangulated surface modeling. The grid created with our technique accurately respects the geometry of the geological structures studied. Layer interfaces (horizons), free-form faults, pinchouts and some unconformities can be taken into account and are faithfully represented within the grid. Starting from initial measured or interpreted data, a surface modeling is performed first. The geological structure modeled is then divided into blocks according to the different layers and the major faults. Then, a grid is generated in each block separately using an adaptation of Coons's interpolations. Finally, the gridded blocks are assembled to build the whole structure. The main issue here is that a block-wise structured curvilinear grid is constructed from triangulated surfaces that best approximate measured data of rather complex geological structures. In addition, the geometry of the main geological features is respected in the gridded structure.
Actually, the initial purpose of the techniques presented was to generate a 3-D mesh of complex 3-D geological structures for geomechanical simulation using finite elements. It turned out that the proposed techniques can be used for stratigraphic grid generation. They are being embedded in Gocad for this purpose, especially for blockwise stratigraphic grid generation within complex geological structures. It also turned out that our approach is general enough to reproduce several kinds of grids used in reservoir simulation. The reason is that all these applications have in common a great need for a good and faithful representation of geological features.