Generating Bit-Rock Interaction Forces for Drilling Vibration Simulation: An Artificial Neural Network-Based Approach

This paper presents a simulation-based artificial neural network (ANN) model to predict bit-rock interaction forces during drilling. Drill string vibration poses a significant challenge in the oil, gas, and geothermal industries, leading to non-productive time and substantial financial losses. This research addresses the challenge of modelling bit-rock interaction excitation forces, which is crucial for predicting vibration and component fatigue life. For a PDC bit with multiple cutters, the cutter tangential velocities at various drilling speeds are calculated, and individual cutter forces are predicted with a two-dimensional discrete element method simulation in which a single cutter moves in a straight line through rock modelled as bonded particles. This data is then used to train an ANN model that characterizes the bit-rock force time series in terms of frequency, amplitude, and distribution of force peaks. Once inserted into a dynamic simulation of the drill string, the algorithm reconstructs the expected bit-rock force time series. A case study using a rigid segment axial and torsional drill string model was used to show that the bit-rock model outputs lead to the expected bit-bounce and stick-slip under certain drilling conditions. Next, the model was implemented in a 3D deviated well drill string simulation with non-linear friction and contact, generating complex stress states with good computational efficiency.