Application of Artificial Intelligence to Model Stresses and Failure Parameters in Anisotropic Formations

Abstract AI-driven solutions have transformed computational applications, offering a competitive edge over traditional analytical and numerical methodologies. While the initial training of these proxy models may be time-intensive, once trained, they execute tasks in almost real-time. This rapid execution is especially beneficial for operations that demand real-time responses, like drilling operations. Rocks generally encountered in oil and gas drilling operations are anisotropic due to their mineral texture, laminations and pre-existing cracks and fractures that develop during depositional and later rock maturing processes. In this work, we perform an in-depth investigation of proxy formulations for analytical Lekhnitskii-Amadei (LA) solution to model stresses around circular cavities in anisotropic rocks. First, we create the forward model by implementing the LA solution for a variety of input parameters on a 2D grid and then post-process the resultant 2D stress states with an iterative algorithm that provides a workflow to estimate compressive and tensile failure limits to establish secure well pressure gradient or mud-weight window. We follow this up with 3 distinct proxy solution approaches that can model stress states on 2D grid, 1D stress vectors around the borehole wall and single output parameters related to LA solution. Specifically, our workflow includes 3 steps. Step 1: We provide a new deep-learning-based reduced-order modeling (ROM) framework for modeling stress distributions around boreholes in 2D grid. Step 2: Realizing the need to model the stresses at the wellbore wall only for most wellbore stability applications, we provide modifications to DL-ROM framework to handle 1D stress vectors at the wellbore wall. Step 3: We then perform a Machine Learning based study to develop proxy solutions for single-value outputs including induced maximum and minimum principal stress at the wellbore, far-field maximum and minimum horizontal stress, shear and tensile wellbore pressure limit for failure as well.