A Digital Twinning Methodology for Vibration Prediction and Fatigue Life Prognosis of Vertical Oil Well Drillstrings

A detailed methodology to develop a digital twin has many useful applications in the era of technology 4.0. This study provides a framework to develop a digital twin for vibration prediction and fatigue life prognosis of a vertical oil well drillstring. The nature and the severity of the down-hole vibrations are identified and estimated based on the vibrational and operational parameter measurements made at the surface level. Because of the difficulty in accessing full-scale industrial drilling data, a reduced-scale drillstring was constructed that could exhibit bit bounce, stick-slip and whirl. A bond graph simulation model was tuned to match the apparatus, and then used to generate synthetic training data. The trained machine learning algorithm can classify the incoming surface monitoring data from the physical twin into different types and severities of vibration states which are not otherwise observable. Moreover, the classified vibration condition is used to re-configure the bond graph model with appropriate complexity to generate a loading history for fatigue life prognosis. The fatigue life estimation uses a novel combination of a low-complexity model of the entire drillstring and a high fidelity finite element model of components where stress concentrations are most severe. The digital twin detected the vibration type and its severity and estimated the remaining fatigue life of the physical system using only measurements of the motor current, rig floor axial vibration, and rotary speed.