OasisLMS

Abstract: Measurement While Drilling (MWD) is increasingly important in geotechnical drilling due to the need for well-informed, real-time decision-making based on data that accurately reflects soil conditions. Measurements such as torque, rotation rate, and fluid pressure hold enormous potential for insights. However, reliably correlating MWD data with geotechnical properties remains challenging due to the complex nature of the signals gathered. This study involves a data-driven approach to address this challenge by applying prob- abilistic machine learning techniques, specifically Bayesian Inverse Modeling, to correlate MWD data with key geotechnical properties. Our approach leverages Bayesian methods to handle inherent uncertainties and variability in the data, offering a robust framework to infer soil and rock properties from real-time drilling pa- rameters. By constructing a probabilistic model, we aim to quantify the uncertainty in geotechnical property estimates, enhancing the reliability of MWD-based stratigraphic profiling and subsurface property determina- tion. We anticipate that this probabilistic approach will enable more accurate and reliable assessments of soil and rock properties, facilitating improved real-time decision-making in geotechnical projects.

Mr. Honglai Peng

International Conference Proceedings