Drilling Memo Analytics for Lost Circulation and Kick Detection

The patent creates a virtual model (Bayesian network) that replicates the functionality of expensive downhole monitoring tools by processing readily available surface data. The system copies the detection capability through software modeling rather than physical downhole sensors, achieving similar detection accuracy without the high cost of downhole equipment.

The patent replaces the mechanical/physical downhole monitoring system with an information-processing system using Bayesian networks and natural language processing. Instead of using physical sensors downhole, the system substitutes a computational model that analyzes surface data patterns to detect lost circulation and kick events.

The patent merges multiple data sources (mud volume measurements, drill string position data, pump status information) into a single integrated Bayesian network model. By combining these previously separate monitoring functions into one unified system, the patent reduces overall system complexity while improving reliability through cross-validation of multiple parameters.

The system implements feedback loops where the Bayesian network continuously compares expected vs. actual measurements and adjusts its predictions accordingly. The natural language processing component provides feedback by classifying drillers' notes to confirm or refute detected events, creating a self-correcting system that reduces false alarms while maintaining manageable complexity.