Session: 01-10 Digital Transformation in Inspection, Compliance, and Engineering Management

Paper Number: 172527

172527 - Driving Robotic Inspection Autonomy Through Multimodal Artificial Intelligence 

Oil and gas production remains a foundational pillar of the global energy landscape, supporting critical industrial and societal infrastructure. Production facilities are intricate systems engineered to manage the treatment, processing, and measurement of oil, gas, and water. These environments rely on essential equipment — including treatment units, motors, and pumps — that require consistent inspection and maintenance to ensure safe, efficient, and sustainable operations. Accurate fault detection in process equipment is imperative, as undiagnosed issues can result in costly repairs, operational downtime, and serious safety hazards. To improve reliability and reduce human exposure to hazardous conditions, the industry is increasingly adopting condition-based and predictive maintenance strategies, leveraging robotic inspection technologies.

Robots equipped with optical cameras, thermal imagers, and microphones can collect rich multimodal data during inspection rounds. Artificial intelligence (AI) is a key enabler in transforming robotic sensor data into actionable insights. In the domain of robotic inspection, multimodal AI plays a transformative role by integrating visual, thermal, and acoustic inputs to deliver real-time, comprehensive intelligence that enhances situational awareness and supports autonomous decision-making. This presentation will highlight our latest advancements in autonomous robotic inspection powered by multimodal AI. Through detailed case studies, we will demonstrate how machine learning-based computer vision and generative AI are being applied to automate routine inspection tasks — such as gauge reading, level monitoring, thermal and acoustic anomaly detection, and corrosion assessment — in live operational environments. These AI models can serve as the compute engine within a cloud-based software platform, providing a scalable, end-to-end digital solution for industrial inspection. This approach promotes sustainability by reducing reliance on on-site personnel, minimizing exposure to hazardous environments, and significantly improving early fault detection, operational reliability, and inspection efficiency.

Presenting Author: Fei Song SLB

Presenting Author Biography: Fei Song currently works as a senior data scientist with SLB, developing AI/ML-based models to enable autonomous solutions.