A probabilistic multiscale modeling for predicting poromechanical properties of shales is presented. To this end, a framework of experimental characterization, physically-based multiscale modeling and uncertainty quantification that spans from nanoscale to macroscale is utilized. To account for the uncertainty in the model input parameters, they are modeled as random variables. To this end, input parameters are divided into two classes of random variables: tensor-valued and scalar random variables and their corresponding statistical description is constructed by employing Maximum Entropy principle (MaxEnt) based on available information. Then, to propagate uncertainty across different length scales the Monte Carlo simulation is carried out and consequently probabilistic descriptions of macro-scale properties are constructed. Furthermore, a global sensitivity analysis is carried out to characterize the contribution of each source of uncertainty on the overall response. Finally, methodological developments are validated against experimental test database. The integration of experimental characterization, multiscale modeling and uncertainty quantification utilized in this work improves the robustness and reliability of predictive models for poromechanical behavior of shales.
展开▼
