Abstract This paper presents a modular, AI-driven platform for real-time drilling performance monitoring, developed to enhance operational decision-making and reduce non-productive time. Built on a microservice architecture, the system integrates both real-time and simulated data sources, including a drilling simulator and a mock application programming interface (API) that replays historical field data as if it were live. This hybrid approach enables controlled, repeatable testing while supporting model development, validation, and demonstration in realistic environments. Continuous learning is achieved through multi-model digital twins that update in real time, fine-tuning pre-trained models without restarting from scratch. Multiple supervised and temporal machine learning models run in parallel, each dedicated to monitoring specific drilling parameters such as weight on bit, rate of penetration, torque, and hookload. A visualization dashboard provides real-time comparison of predicted and measured values, along with anomaly detection based on deviation thresholds and trend shifts. Large language model (LLM) agents are incorporated to process drilling reports, extract structured parameters, and support natural language queries through a chat-based interface. Demonstrations using simulator data and replayed field datasets confirm the system's capability to integrate diverse AI components for monitoring and scenario testing. These results indicate that the proposed architecture provides a flexible and scalable foundation for deploying AI-enhanced drilling performance systems in operational environments, with potential to accelerate adoption of advanced analytics in the drilling industry.
Sahebi et al. (Mon,) studied this question.