Neural Network Model to Predict Pore Pressure and Wellbore Stability from LWD Log and Drilling Parameter

Abstract

ABSTRACT In well construction, well stability is needed to predict drilling time, costs, and risk management accurately. Several models have been developed in the last 60 years to estimate rock parameters and stress magnitudes from well logs. However, this involves uncertainties due to rock heterogeneity, lack of information due to high data acquisition costs, and indirect measurements. Drilling monitoring has been used to mitigate instability occurrences by updating the geomechanics model while drilling. Nevertheless, monitoring highly depends on the analyst's experience and immediate interpretation of available information. Artificial Intelligence algorithms have been used to improve efficiency in many different technologies, but they depend on good predictions to learn from experience. In order to build an accurate and generalist ANN model, this paper uses the real case study experience of the analysis of offshore wells in Brazil calibrated with drilling events. Data is provided by the Brazilian National Petroleum Agency (ANP) and used to build an ANN model. The available drilling parameters are used, such as weight on bit, torque, rate of penetration, log data and well reports. Neural networks were applied to the available data to build a model. Model output includes pore pressure and collapse pressure for the survey. The proposed model can be an additional tool to help well construction operations. INTRODUCTION The maximum value of pore or collapse pressure defines the lower boundary of the drilling mud window. Using mud weights below those values can lead to several problems, such as inflows, excess reaming, poor hole cleaning conditions, excess cavings, pack-offs, stuck pipe, and more severe issues, such as loss of the hole or uncontrolled blowouts. Methods to estimate pore pressure, in-situ stresses, and collapse curves were developed and improved over many years (Barton et al., 1997a; Eaton, 1972, 1975a, 1976; Fjaer et al., 1992; Zoback et al., 1985; Zoback, 2007). Most pore pressure methods are applied to argillaceous rocks assuming under-compaction mechanisms (Vernik, 2016). Pore pressure in permeable environments is generally measured using well pressure tests or inflow events.