Analysing the Effects of Different Pump Start-Up Sequences on Wellbore Stability

Downhole pressure management is critical to maintaining wellbore stability, particularly in deep-water and narrow-margin drilling environments. A key contributor to pressure fluctuations is the breakdown of gelled drilling fluid structures formed during static periods. When circulation resumes, additional pump energy is required to break the gel, and poor pump start-up practices can lead to sudden annular pressure spikes that threaten well integrity. Therefore, understanding and predicting downhole pressure changes while breaking down the gelled structure during pump restarts are of paramount importance. Despite numerous studies conducted in the past to investigate the relationship between pump start-up sequences and associated pressure changes, all of these studies utilized only surface drill data, neglecting downhole data for various reasons. This thesis investigates the relationship between pump start-up sequences and annular pressure loss by analyzing historical real-time surface and downhole data from multiple wells across different regions. Unlike previous studies that relied solely on surface data, this work incorporates annular pressure loss measurements to provide a more accurate understanding of downhole dynamics. To perform a comprehensive analysis and reach a sound conclusion, eight new performance benchmarks were developed in this thesis. In addition, the thesis aims to develop empirical and data-driven models that can be used to predict the magnitude of fluctuations in annular pressure loss during pump startup. The main findings of the thesis are as follows: The most critical parameters causing significant fluctuations in annular pressure loss are the incremental change in flow rate, connection time, and the annular space between the drill pipe and the wellbore. Among the five models examined in this thesis to predict the magnitude of fluctuations in annular pressure loss, Random Forest and polynomial regression demonstrated the highest predictive performance.