Advanced fluid dynamics in well construction with application of hybrid physics-based and data-driven modeling

Modeling of the physical systems is essential to capture action and response relationships during well construction processes, such as while drilling for geothermal, oil and gas wells. Accurate fluid dynamics models are required especially to fully automate the well construction process, which will significantly improve the safety, security and consistency of the operations. However, it is difficult to achieve robust and accurate modeling of any inherently complex drilling process. Recent publications across various domains showed that data-driven models have the potential to capture complex relationships successfully. However, purely data-driven approaches come with characteristic disadvantages. For instance, they lack interpretability and struggle to properly capture causal relationships. This dissertation focuses on three major fluid dynamics research topics in well construction: cuttings transport, non-Newtonian fluid flow in drilling circulation systems and thixotropic fluid behavior. Specifically, physics-based modeling of each topic is investigated and then several hybrid modeling applications are attempted through combining physics-based and data-driven models via rule-based stochastic decision-making with the goal of mitigating the disadvantages of using the two types of approaches standalone. Furthermore, the last chapter demonstrates how better fluid dynamics modeling will lead to more robust control of rig actuators. This dissertation aims to advance fluid dynamics research and open the way for further hybrid modeling applications in well construction.