Inside storage tanks and pressure vessels, access is tight, visibility is limited, and the risks are well known - low oxygen, toxic vapors, and, in some cases, explosive atmospheres. These are permit-required confined spaces, and despite strict procedures, they remain one of the most dangerous parts of refinery operations.
The industry has accepted this trade-off for years: critical inspections, carried out in high-risk conditions, because there was no practical alternative. What’s changing now is that there finally is one. Robotic systems are starting to take on these inspections, removing the need for human entry while delivering more consistent and repeatable data.
Save this PDF for later by downloading it here.
The Scale of Inspection in Refineries
Once you step back and look at a refinery as a whole, the scale of inspection becomes hard to ignore. A single site can have hundreds, sometimes thousands, of assets that need routine checks under standards like API 510 for pressure vessels and API 570 for piping.
Traditionally, that means internal inspections. A certified inspector enters the vessel, documents corrosion, cracking, coating breakdown, and weld defects, then exits so the system can be cleaned, purged, and brought back online. It’s a process that’s tightly controlled, but also slow, resource-heavy, and disruptive to operations.
The downtime alone is a major issue. Preparing a tank for inspection (cleaning, isolating, and making it safe to enter), can take days or even weeks. During that time, the asset isn’t generating revenue. Add in permits, specialist personnel, and safety infrastructure, and the cost quickly adds up.
But the bigger issue is risk. Confined space incidents continue to account for a disproportionate number of fatalities in the oil and gas sector. And in many cases, it’s not the initial entrant, but the rescue team that ends up exposed.1,2
Robots remove the need for the whole process around confined space entry.
Crawler Robots for Internal Inspection
Crawler robots are already replacing the need for confined space entry in refineries.
They move on tracks or wheels across the floors and walls of tanks, vessels, and pipelines, carrying cameras, ultrasonic thickness gauges, and corrosion sensors. That allows operators to inspect asset condition without sending anyone inside.
Some systems can run routine inspections with minimal input, with operators stepping in when needed. A recent study in Scientific Reports described a modular crawler built around a Raspberry Pi 4 for pipeline inspection. The design allows sensor units to be swapped depending on the job, which is useful in refineries where pipe sizes and conditions vary.
This is already being used at scale. Companies like Dow are deploying crawler-based systems alongside providers such as Eddyfi Technologies, with the aim of reducing confined space entry wherever possible.3,4
Using Drones Inside Tanks and Vessels
Crawlers work well where they can stay in contact with a surface. The limitation is everything they can’t reach.
That’s where aerial systems start to play a part. Unmanned aerial vehicles (UAVs) move through the full volume of a tank or vessel, rather than being restricted to floors and walls. That changes how inspections are carried out, especially in large or complex structures.
Collision-tolerant drones, often fitted with protective cages, can operate in tight, obstacle-filled spaces like floating-roof tanks. They capture high-quality visual, thermal, and LiDAR data while navigating around internal structures that would slow down or block ground systems.
In practice, this means faster inspections with less setup. Drones equipped with thermal imaging and gas detection sensors can collect real-time data, supporting predictive maintenance and early leak detection without requiring human entry. A trained pilot can complete a visual inspection of an external storage tank in around 75 minutes, allowing multiple inspections to be carried out in a single day.
Regulatory acceptance has also increased, with drone data meeting API and OSHA standards when quality and coverage are sufficient.5,6
Navigation Without GPS in Confined Spaces
One major challenge, however, is that none of these systems can rely on GPS once they’re inside.
Refinery vessels are difficult environments to navigate, with metal walls, curved surfaces, low texture, and no external positioning signals. For a robot or drone, this makes it hard to know exactly where it is or how to move reliably through the space.
But SLAM (Simultaneous Localization and Mapping) offers a solution in this regard. Instead of relying on external signals, the system builds a map of the environment in real time while tracking its own position using onboard sensors.
In practice, that usually means combining data from cameras, LiDAR, and inertial measurement units (IMUs). Research has shown that tightly integrated, multi-sensor systems perform much better in these conditions, especially when the data is fused at the sensor level rather than processed separately.7,8
There are still limits. Processing large volumes of data in real time can be demanding, and sensor drift can build up over longer inspections. But with proper calibration and hybrid sensor setups, these systems are now capable of navigating large storage tanks independently, generating mapped inspection data and flagging areas of concern without relying on external positioning.
Non-Destructive Testing on Robotic Platforms
Being able to see the inside of a vessel is one thing. Understanding its condition is another.
Many of the most critical issues in refineries, like corrosion, wall thinning, or internal defects, aren’t always visible. Non-destructive testing (NDT) can be deployed here to take these measurements directly, without requiring human contact. Robotic systems are now being equipped with NDT tools to assess material condition in detail while the robot is still inside the asset.
One area getting particular attention is electromagnetic acoustic transducers (EMAT). Unlike conventional ultrasonic methods, EMAT doesn’t require surface preparation or liquid couplants, which makes it far more practical in dirty or wet environments, conditions that are common inside tanks and pipelines.
Other NDT methods, such as robotic ultrasonic testing, magnetic flux leakage (MFL) scanning, and radiographic testing, have also been successfully used on robotic crawlers for inspecting pressure vessels. Each method targets a different type of defect.9,10
This shift is also reflected in the market.
The global NDT market in oil and gas was valued at $9.9 billion in 2022 and is projected to reach $19.2 billion by 2030, with robotic and automated crawlers cited as a primary driver of that growth. The API formalized this trajectory in May 2023 by signing a memorandum of understanding with the American Society for Nondestructive Testing to promote quality NDT standards across the oil and gas sector. This move implicitly acknowledges robotics as part of the compliance framework going forward.11
AI and Digital Twins for Inspection Data
A single robotic inspection can generate thousands of ultrasonic readings and hundreds of high-resolution images in one pass. That’s far more than a human inspector can realistically review in detail. Artificial intelligence (AI) and machine learning (ML) solve this throughput problem via automated defect detection and classification.
Deep learning models trained on corrosion images can now accurately identify and categorize surface anomalies, achieving performance comparable to that of experienced inspectors. Additionally, convolutional neural network (CNN)-based systems have shown reliable detection of oil levels and defect identification from sight-glass images, even under varying lighting and operational conditions.12
On top of that, digital twin platforms provide a way to organise and track all of this information over time. By creating a three-dimensional model of the asset and linking inspection data to specific locations, engineers can monitor how defects develop across inspection cycles and make more informed decisions about maintenance and remaining service life.
Baker Hughes and similar service providers now offer Robotics-as-a-Service (RaaS) models that bundle the robot, sensor package, AI analysis pipeline, and digital twin platform into a subscription service, which makes this capability accessible to refineries that lack the capital or expertise to develop these systems in-house.13
What Lies Ahead?
For a long time, the assumption was simple in that, if you needed to inspect something, someone had to go in.
But that mindset is starting to look outdated.
The tools are already doing the work in places where access used to slow everything down or make it risky. Not everywhere, and certainly not perfectly, but enough that the old way of planning inspections doesn’t quite hold up in the same way.
What comes next is less about capability and more about how this gets built into everyday practice. That means integrating these systems across sites, getting inspection data into formats regulators are comfortable with, and making sure the results are trusted when decisions are being made.
There are still gaps. Higher temperatures and harsher conditions are a challenge, and the systems need to hold up better there. But that’s already being worked on.
In the meantime, the direction is fairly clear. Companies are trying to reduce, and in some cases remove, the need for confined space entry altogether. The question now is, how quickly can that become the default?1,2
References and Further Reading
- Thirunagalingam, A. et al. (2025). Application of Automation and Robotics in the Oil and Gas Industry. Revolutionizing AI and Robotics in the Oil and Gas Industry, IGI Global. DOI:10.4018/979-8-3693-8156-4.ch011, https://www.igi-global.com/chapter/application-of-automation-and-robotics-in-the-oil-and-gas-industry/376729
- Robotic Inspection Inside Confined Spaces. (2021). Nexxis. https://nexxis.com/robotic-inspection-inside-confined-spaces/
- Eltwab, A. A., & Sameh, A. (2026). A modular, multi-sensor crawler robot for adaptive pipeline inspection: Design and experimental validation. Scientific Reports, 16(1), 880. DOI:10.1038/s41598-025-32719-y, https://www.nature.com/articles/s41598-025-32719-y
- Johnson, P. (2022). Access Granted with Confined Space Entry Inspection Robotic Crawlers. Eddyfi. https://blog.eddyfi.com/en/access-granted-with-confined-space-entry-inspection-robotic-crawlers
- Nooralishahi, P. et al. (2021). Drone-Based Non-Destructive Inspection of Industrial Sites: A Review and Case Studies. Drones, 5(4). DOI:10.3390/drones5040106, https://www.mdpi.com/2504-446X/5/4/106
- Drones for Inspection: Advantages, Use Cases and Best Drone Companies. Voliro. https://voliro.com/blog/best-drones-for-inspection/
- Tripicchio, P. et al. (2018). Confined spaces industrial inspection with micro aerial vehicles and laser range finder localization. International Journal of Micro Air Vehicles, 10(2), 207–224. DOI:10.1177/1756829318757471, https://journals.sagepub.com/doi/10.1177/1756829318757471
- Olleik, H. et al. (2025). Challenges and Advances in SLAM-Based Inspection for Low-Texture and Confined Environments: A Systematic Review. Research Square. DOI:10.21203/rs.3.rs-6751721/v1, https://www.researchsquare.com/article/rs-6751721/v1
- Tian, Y. et al. (2024). Non-destructive testing technology for corrosion wall thickness reduction defects in pipelines based on electromagnetic ultrasound. Frontiers in Earth Science, 12, 1432043. DOI:10.3389/feart.2024.1432043, https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2024.1432043/full
- Ni, S. (2024). Study of Robotic Non-destructive Testing Solutions. Highlights in Science, Engineering and Technology, 106, 129–136. DOI:10.54097/x5kkm090, https://drpress.org/ojs/index.php/HSET/article/view/22800
- Non-Destructive Testing Market - 2024-2031. (2026). Data M Intelligence. https://www.marketresearch.com/DataM-Intelligence-4Market-Research-LLP-v4207/Non-Destructive-Testing-44409216/
- Abbas Al-Jiryawee. (2026). Autonomous Robotic Inspection System for Oil Tank Level Detection Using Deep Learning and Smart Vision Sensors. Rafidain J. Eng. Sci., vol. 4, no. 1, pp. 28–44. DOI:10.61268/x1542751, https://rjes.iq/index.php/rjes/article/view/303
- Robotic inspection more accessible than ever. Baker Hughes. https://www.bakerhughes.com/sites/bakerhughes/files/2023-06/way-raas_industrybrief-flight1-r2.1.pdf
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.