Current methods for remote visual inspection of confined spaces, such as pressure vessels, reactors, and boilers, typically involve the use of human-operated robots, drones, or pole-mounted camera systems. These inspections often rely on manual data collection, followed by report generation in which selected images and data are inserted into predefined templates. This process usually lacks a direct connection between the inspection data and the precise location within the asset, resulting in additional manual effort to integrate the findings into internal workflows. Consequently, the value of the data in supporting digital strategies for asset owners and operators is diminished.
For certain asset types, inspection procedures and recommendations provided by organizations such as API, ASME, and HOIS offer guidance on what to inspect and how to perform the inspection. While these standards support planning and execution, the responsibility for ensuring that all relevant features are properly examined typically rests with the inspector. Maintaining image quality in accordance with industry codes and standards also remains a persistent challenge.
Part one of the presentation covers recent advancements in robotic inspection technologies. These include 3D dynamic reconstruction for real-time digital twin creation, the embedding of inspection data within asset models, and the use of AI for detecting and extracting relevant objects and features. AI can assist inspectors during the inspection process or enable substantial automation. These innovations directly address the key challenges of confined-space inspections described above. The presentation also underscores the importance of localization technologies, such as lidar and 3D modeling, for accurate navigation, and demonstrates their application across mobile robotic crawlers, drones, and pole-mounted camera systems. Additionally, it provides best practices for ensuring image quality during robotic inspections to meet industry standards.
Part two of the presentation introduces a concept for integrating robotic and AI capabilities into a pole inspection camera system. This approach is illustrated through examples from inspection missions across industries such as Oil & Gas, Chemicals, and Power Generation.
To validate the system’s capabilities, a comprehensive evaluation was performed using a test vessel. All inspection data was geotagged and incorporated into a real-time digital twin generated via 3D dynamic reconstruction. The inspection focused on critical features including nozzles, supports, and welds, with each image precisely mapped within the 3D model. Using AI, key inspection targets were automatically identified, enabling the system to autonomously navigate from one feature to the next.
The results, benchmarked against a traditional manual inspection performed by an inspector physically entering the vessel, confirmed full compliance with applicable inspection standards. The digital twin significantly enhanced data management and post-inspection analysis. Furthermore, automated report generation substantially reduced the time and effort required for documentation.
The presentation concludes by sharing key insights from the evaluation and providing practical recommendations for integrating robotics into routine inspection operations