Improving Drilling Simulation Computation Performance with Smart Logic and Machine Learning

Abstract

Modeling and simulation play a key role in well construction planning, which can help to evaluate and optimize the engineering designs for a well. Today, many simulations use finite element analysis (FEA) and computational fluid dynamics (CFD) to model complex dynamic downhole conditions and behaviors of drilling tools. However, one challenge is that the complex simulation may take a few hours to run, which limits the usage to only a few well planning jobs. This limitation also poses as barrier in real-time monitoring applications, where under one second computation speed is required. In this paper, two approaches are presented for improving the performance of drilling simulations: smart depth selection logic for BHA tendency calculation, and reduced order model using machine learning for motor optimization modeling.