Abstract
Rock mechanical properties, such as the Poisson's ratio, Bulk, Shear and Young's moduli are required for various oilfield operations such as seismic processing, reservoir simulation studies, and hydraulic fracturing for sand control. These properties are usually determined from compressional (P) and shear (S) wave velocities which can be captured by the dipole sonic logging tool. For unconsolidated rocks (shallower than 6000 ft), the P and S-wave velocities are difficult to predict due to the dispersive nature of these rocks. Also, the available correlations in the open literature to determine these parameters should be used with caution since these were developed mainly for deeper formations.
In this study, we compared the sonic wave velocities empirically predicted by Castagna et al., Eberhart-Phillips and Han with the logged sonic wave velocities across the Cruse Formation for eight onshore wells in South-Western Trinidad. We also investigated more suitable methods of matching the logged velocities across the Cruse, by developing new empirical models and an analytical method (Hertz-Mindlin contact model) to account for the unconsolidated rocks. The basic input data required for predicting the sonic velocities and dynamic rock mechanical properties were porosity, clay content and bulk density which were taken from the log data. The models for obtaining the static rock mechanical properties were tested against the results of Mini-Frac tests in two offset wells in the Cruse.
The Artificial Neural Network (ANN) empirical method was shown to be best suited for predicting the dynamic rock mechanical properties for the Cruse Formation, since Average Absolute Errors (AAE) of 3 and 6 % were obtained in matching the P and S-wave velocities respectively. Similarly, the AAE obtained from applying the three empirical correlations were worse than 17 and 30 %. This error was generally observed to lessen for wells which encountered the Cruse at greater depths. While AAE of 6 and 12 % were obtained from applying the analytical method, reservoir depletion was found to significantly affect the results of the modeling due to the implied presence of hard-to-predict free gas saturations. Also, it was found that the analytical method is the only viable procedure to predict the static rock mechanical properties in lieu of laboratory core tests. The analytical method has been tuned in this study using True Vertical Depth (TVD) as a correlation parameter for the first time, and it can generally be used to estimate both the dynamic and static rock mechanical properties of the Cruse Formation for future oilfield operations.