A UKF enabled model based event detection system for drilling operations

Jiamin Xu

Abstract:

This paper presents model-based event detection systems integrated with an Unscented Kalman Filter (UKF) for
drilling operations. The key novelties include a refined mathematical model, a UKF-based event detection sys-
tem, and phase portraits for event analysis. The paper first presents a refined mathematical model designed to
enhance the prediction of frictional pressure losses in drilling operations, accommodating both laminar and
turbulent flow conditions in non-Newtonian fluids and considering the impact of directional flow. The model’s
accuracy and reliability are confirmed through comparisons with existing datasets and experiment. Then,
building on this validated model, this paper introduces an observer-based event detection system that is based on
the UKF and compares the UKF with a conventional adaptive nonlinear observer (ANO). A detailed comparison
of these two observers assesses their effectiveness in detecting specific drilling events, such as gas kicks and pack
offs. Comparison and validation with two datasets demonstrates the UKF’s reliability in event detection. Finally,
this paper discusses phase portraits that depict various drilling events, enhancing the understanding of these
occurrences beyond the existing literature and suggesting the potential use for recognizing and responding to
unexpected events. This analysis extends the detection system’s applicability and utility across various drilling
scenarios, highlighting its importance in improving operational safety and efficiency.