Machine Learning Enhanced Upscaling of Anisotropic Shear Strength for Heterogeneous Oil Sands

Abstract Weak shale beddings are widely distributed in the overburden and reservoir of oil sand deposits and lead to reduced anisotropic shear strength. Understanding the shear strength of the overburden and the reservoir is important in risk assessment of slope stability in open-pit mining and caprock integrity of in-situ thermal recovery of oil sands while optimizing the production of bitumen. Due to the restrictions of computational efficiency, cells used for simulation cannot be smaller enough to capture the details of heterogeneity in the reservoir. Therefore, a robust and efficient upscaling technique is important for modeling the impact of heterogeneity on the deformation and failure of oil sands during mining and in-situ recovery. However, current analytical and numerical upscaling techniques cannot provide computationally efficient geomechanical models that consider the impact of inclined shale beddings on shear strength. Therefore, we propose a machine learning enhanced upscaling (MLEU) technique that leverages the accuracy of local numerical upscaling and the efficiency of machine learning techniques. MLEU generates a fast and accurate machine learning-based proxy model using an artificial neural network (ANN) to predict the anisotropic shear strength of heterogeneous oil sands embedded with shale beddings. The trained model improves accuracy by 12%-76% compared to traditional methods such as response surface methodology (RSM). MLEU provides a reasonable estimate of anisotropic shear strength while considering uncertainties caused by different configurations of shale beddings. With the increasing demand for regional scale modeling of geotechnical problems, the proposed MLEU technique can be extended to other geological settings where weak beddings play a significant role and the impact of heterogeneity on shear strength is important.