A spatiotemporal data-driven homogenization approach for hierarchical modeling of multiphase frozen soil in permafrost

点击翻译按钮获取中文摘要

Global issues related to extreme climate change and demanding energy resources have created new engineering problems. Those issues are associated with sustainable development and resilient infrastructure in our society, such as heavy rainfall-induced slope stability, sinkhole due to groundwater level change, hydraulic fracturing for unconventional energy recovery, geothermal energy facility, nuclear waste disposal, etc. Within the Canadian context, the melting of frozen soil is a critical issue in permafrost. The main reason is global warming that accelerates the thawing and subsequently the cyclic freezing, which in turn results in damaging buildings and infrastructures. In other words, frozen soils in permafrost are subjected to coupled loading conditions in terms of thermal, hydraulic, and mechanical aspects. As a mixture of air-water-ice-solid, furthermore, the multiphase frozen soil possesses multiscale characteristics that show size-dependent responses. Therefore, the thawing and freezing soil in permafrost requires our fundamental knowledge and advanced technologies to understand, analyze, and predict its complicated behavior. In the short term, the proposed research program is designed to develop a unified computational framework that transcends the multiscale behavior of multiphase frozen soils based on a hierarchical modeling approach. Focus Area I concentrates on deriving theoretical formulations to characterize the coupled thermo-hydro-mechanical processes under freeze-thaw actions for the pore-scale level. Focus Area II involves developing a data-driven constitutive model aiming at specimen-size level by leveraging machine learning to improve multiscale computational cost and flexibility. The pore-scale simulations from Focus Area I will be used to generate a database for training the data-driven model. Focus Area III involves constructing a unified framework for field-scale simulations by developing a multiscale bridging algorithm. The domain mapping technique to accommodate the spatiotemporal information of frozen soil will be further developed to be combined into the framework. This process will integrate the achievements from Focus Areas I and II and field/experimental data - a hybridized modeling method. In the long-term, the unified computational framework will be further advanced to interactively update monitoring information or additional site investigation in permafrost. This hybridized computational framework will provide an innovative methodology in analyzing conventional engineering problems, validating designs, and making predictions for resilient infrastructures, intelligent energy systems, and challenging environmental issues in Canada. The ultimate goal and strategic plan of the applicant's research will further provide cutting-edge knowledge and advanced technical skill set not only to train High Qualified Personnel (HQP), but to support industry, government, and academia in Canada and across the world.