Print this pageNew Techniques for Fracture Distribution and Prediction from Kinematic Modelling of 3D Strain Fields
GIBBS, ALAN D., FAISAL JAFFRI, and TITUS MURRAY
The traditional approach to predicting fracture distribution in reservoirs involves a combination of either curvature analysis and or statistical extrapolation of fracture information from well or borehole logging. These two techniques are inadequate in that they take insufficient account of the geological development of the structure. Curvature analysis is based on a simple buckle fold model which takes no account of fold hinge migration with time, for example in fault bend or roll-over folding. A purely statistical approach on the other hand cannot be used to predict distributions on a geological basis or to discriminate between different fracture sets or timing.
For these reasons a new approach has been developed, based on kinematic forward modeling of the structure. The technique involves constructing a detailed 3D model of horizons and fault linkages using seismic and borehole information. The model is then restored to its pre-faulting, pre-deformation configuration. The restoration in 3D provides a validation of the interpretation from which the solution space and confidence limits can be determined for both the restoration and interpretation parameters.
The pre-deformation geology is subsequently populated with strain markers and forward modeled to its present day configuration using the optimum restoration parameters previously determined from the solution space analysis. The process allows for both finite and incremental 3D strains to be calculated. Using the incremental strains and associated strain trajectory maps it is possible to discriminate between fracture sets generated at different stages in the structural evolution. As the incremental strain ellipsoid is coaxial with the stress ellipsoid at that stage it is possible to make predictive models for paleo-stress analysis.
This paper describes the techniques used in this new approach, and outlines some of the preliminary results using this analysis on a well constrained data set.