A performance-driven surveillance architecture for anomaly detection and response in subsea systems using PI Processbook and AI models

This paper presents a performance-driven surveillance architecture that enhances anomaly detection and operational decision-making in subsea oil and gas systems through the integration of PI ProcessBook and artificial intelligence (AI) models. Subsea operations generate vast amounts of time-series data from distributed sensors monitoring pressure, temperature, flow rates, and system states. While PI ProcessBook provides a widely adopted platform for visualizing and analyzing such data, its functionality is primarily limited to human interpretation and threshold-based alerts. To address the limitations of manual monitoring and improve responsiveness, this research proposes a unified architecture that embeds AI-based pattern recognition within the PI ProcessBook interface. The proposed system includes structured data acquisition pipelines, real-time preprocessing, and the deployment of supervised and unsupervised AI models for identifying deviations from normal operating conditions. These models continuously monitor telemetry and feed anomaly insights back into PI ProcessBook in the form of graphical alerts and contextual diagnostics. The architecture supports operator decision-making by reducing information overload, enhancing detection accuracy, and improving response time. Analytical insights demonstrate improvements in anomaly sensitivity, false positive reduction, and operator confidence. The integration also supports predictive maintenance strategies and contributes to digital transformation in offshore production environments. Recommendations for future development include adaptive learning, system-wide data fusion, and autonomous response capabilities. Keywords: Subsea Surveillance, PI ProcessBook, Anomaly Detection, Artificial Intelligence, Performance Monitoring, Predictive Maintenance.