Optimizing Underbalanced Coiled Tubing Drilling Monitoring Via Advanced In-Line Sensing AI Framework

Abstract

Abstract Logging while drilling (LWD) technologies have been extensively developed and implemented to evaluate downhole rock and fluid properties. Nevertheless, sparsity of information, indirect measurements, and a challenging downhole environment are critical factors in obtaining accurate formation evaluation diagnostics. This is especially critical given the limited number of LWD technologies for underbalanced coiled tubing drilling (UBCTD). As such, forecasting formation patterns and drilling conditions from real-time surface in-line measurements on the return mud is a promising area that will significantly advance UBCTD monitoring capabilities. In this work, we present a noninvasive data driven approach by utilizing a series of inline sensors on the mud return to determine the characterization of fluids and rock cuttings. An advanced nonlinear autoregressive network was utilized for the forecasting of the solid and hydrocarbon concentrations based on the sensor derived in-pipe measurements. The measuring sensors comprise a range of inline sensors, including ultrasonic acoustic sensors, optical imaging devices, pipe resistivity sensors, spectral gamma ray sensors, flow meters, and viscosity sensors. The results demonstrate the strong forecasting capabilities of the nonlinear autoregressive network deep learning framework in determining solid and hydrocarbon concentrations and may support classifying rock cutting' types in real-time based on the measurement dataset. The test was conducted on synthetically realistic formation drilling data, which showed strong estimation capabilities and a high-performance score. Based on the test, we concluded that gamma rays, electromagnetic induction, optical, and acoustic sensing are the major principal components affecting the quality of the estimation. This deep learning framework is instrumental in the integration of LWD, measurement while drilling (MWD), advanced mud logging, and drilling parameter data to guide UBCTD operations and support real-time formation evaluation.