April 2023

Ensuring safe operations on subsea wells using machine vision

Having successfully used machine vision to ensure safe well operations, 4Subsea is now looking at how best to harness the technology when it comes to monitoring subsea assets.
Harald Holden / 4Subsea

Based on an industry need to monitor growth in subsea wells, to ensure safe rig operations, 4Subsea developed a subsea camera with machine vision. The solution allows operators to monitor temperature and pressure-driven growth of the subsea wellhead, so that this information can be used when assessing the structural integrity of the well before and during rig operations. 


A subsea well and wellhead system is made up of concentric pipes, all hung off the wellhead. One of its tasks is to transfer bending loads from the riser and blowout preventor to the conductor and then template structure, if present. This job is typically done through the contact in shoulders or support rings in the wellhead system.  

When exposed to pressure and temperature loads, the different pipes in the wellhead system can experience vertical movement relative to each other. If the wellhead lifts relative to the conductor, the intended load transfer between the wellhead and the conductor can be compromised, thereby increasing fatigue loading in the wellhead, with potential significant impact on the wellhead’s load-bearing capacity. 

Normally, the wellhead is prevented from vertical movement, relative to the conductor, through a snap ring that will lock into the conductor when the wellhead is landed. However, on some wells, issues during installation of the well introduce uncertainty around whether the snap ring has been properly engaged.  

For operators, it is critical to determine whether they have wellhead growth relative to the conductor, in order to know the wellhead's load carrying capacity and ensure safe operations.  

In an ROV picture of the wells ahead of the operation, seen in Fig. 1, a connector can be seen on top of the wellhead. Well growth can be observed as an increased distance between the wellhead connector and the top of the template structure. In other words, the amount of well growth can be monitored by measuring the connector gap.  

Fig. 1. The well that was monitored Image: 4Subsea.
Fig. 1. The well that was monitored Image: 4Subsea.


This gap, however, would not differentiate between the following scenarios:  

  • Conductor-template movement: relative movement between the conductor housing and the template structure. 
  • Wellhead-conductor movement: the relative movement between the conductor housing and the wellhead housing. 

For the client, it was important to differentiate between the two situations, as the consequence of the fatigue damage of the wellhead extension weld was significantly different for the two cases. Therefore, a measurement system was subsequently used as a means of differentiating. 

The well growth of subsea wells is a slow process, and pre-project estimations of a vertical expansion of between 0 and 150mm were expected but could have taken place over anywhere from several hours to several days. In addition, the acceleration during the process would be in the range of 10-6 m/s2, well outside the range of traditional accelerometer-based monitoring. 

The chosen camera-based approach with machine vision provided several key benefits:  

  • The actual distance could be measured, allowing for differentiation between the two scenarios.  
  • The solution is easy to retrofit. In many cases, it can be challenging to retrofit sensors on subsea structures, as the position of interest can be difficult to access, using an ROV. A camera, on the other hand, can be placed somewhere which is more easily accessible. 
  • Pictures or time-lapse video, when available, allow for understanding measured results, communicating the findings and handling a measured response that deviates from the expected results. 


In 2016, 4Subsea began the development and testing of different camera-based monitoring technologies, with two versions specifically tested: 

  • Marker-based tracking—where pre-made, easily detectable patterns are attached to the structure in question and used to track position. 
  • Marker-less (or feature-based) tracking—where features in the picture are used to track position.  

A challenge that had to be overcome is that when using cameras for distance-based monitoring, a picture is a 2D representation of what is a 3D space, and so depth (in the picture) is difficult to judge. A marker-based approach was adopted to solve this problem, as a marker has a known size and shape, and the position and orientation in space can be calculated from its 2D shape. 

Another issue was the interference of wildlife, as seen in Fig. 2. The seabed is rich with marine creatures, and one issue was that fishes and starfishes could cover markers, leading to the failure of the marker-based algorithm and periods without measured results. While not a major problem on this particular project, a future solution could be to use a feature-based algorithm to fill in the holes where the marker-based algorithm fails. 

Fig. 2. Wildlife covering the marker. Image: 4Subsea.
Fig. 2. Wildlife covering the marker. Image: 4Subsea.


For this project, two battery-powered camera kits were mobilized, with each consisting of a subsea camera, battery kit with ROV switch (battery life ca 4-6 weeks), flash and brackets with magnets, to fit to subsea template structure. Also used were two target plates with ArUco markers—synthetic square markers, each composed of a wide black border and an inner binary matrix that determines its identifier (id), with magnets and fishtail for ROV handling, Fig. 3b. 

The camera kits, as seen in Fig. 3a, were launched, using a winch and guided into place using a subsea ROV. Over five months, they moved between different wells, with the equipment set up to take two pictures per hour. To recharge the batteries and download the pictures, the camera was retrieved to surface every four to six weeks. Camera calibration, using a standard chessboard, was undertaken in a seawater tank each time the camera was topside. 

Fig. 3a and 3b. The monitoring kit used and magnetic target plates with markers. Image: 4Subsea.
Fig. 3a and 3b. The monitoring kit used and magnetic target plates with markers. Image: 4Subsea.

From the images, as seen in Fig. 4, information of the well pressure, temperature and production was included from the topside production facility. Significant wellhead lift of 55cm was observed on one of the wells monitored. The growth itself started around eight days after the production was turned on, indicating that a certain amount of vertical load is required to overcome the resistance toward well growth. 

Fig. 4. Pictures of the wellhead lift. Image: 4Subsea.
Fig. 4. Pictures of the wellhead lift. Image: 4Subsea.

When the well temperature increased further, the growth correlated, before the conclusion of production when the temperature dropped, and the measured distance dropped within one hour—back to the original position. 

A major benefit of the camera monitoring is the capability to extract further, more detailed information, which in this case allowed for the discovery that the conductor did not move relative to the template. It was subsequently concluded that the measured wellhead lift was due to relative movement between the conductor and wellhead housing. 

For the client, this allowed them to establish the fatigue status of the production wells, which is used to determine their remaining life service. When this is exceeded, the well has to be shut down. 


The initial monitoring campaigns used an off-the-shelf, rental, subsea camera package.  During the project, a decision was made to develop a new, more suitable camera, using the experience from the monitoring phase.  The following key weaknesses of the off-the-shelf camera package were identified: 

  • Camera control—the off-the-shelf camera had built-in features, such as auto zoom and gyro stabilization of the lens. Changes in the focal settings or movement of the lens introduced unnecessary inaccuracy in the measured distance for time-lapse type monitoring.  
  • Battery life—the typical monitoring campaign was three to five months long, while the battery life of the off-the-shelf camera was four to six weeks.  This led to several costly offshore trips for technicians to retrieve and charge the camera. 
  • Labor intensive—the weight of the camera kit was well above 100 kg. The kit had to be run by winch, using the subsea ROV to guide the camera into position. This complicated the operation of the camera, as both winch operators and the ROV crew had to be available to run the camera. 
  • Subsea communication—the off-the-shelf camera was completely autonomous, and no communication was possible with the camera after leaving the workshop topside. Each time camera settings had to be changed or test pictures were needed, the whole kit had to retrieved. Retrieval and deployment were 24-hour operations, for which both the ROV crew and the winch operator had to be available. 

The new camera mitigated all these weaknesses. The camera control settings were optimized for time lapse usage. Battery life was increased to more than nine months, and both the camera and battery’s weights were significantly reduced, so that the running or retrieval could be done by ROV alone. Wireless communication through the SMS Magic Hand was also added, enabling camera settings to be changed and pictures downloaded while the camera is still subsea. 

The new camera was developed and used for the last four monitoring campaigns (see the new installed camera in Fig. 5). The new camera was successfully operated for all campaigns, significantly reducing the cost and complexity of the monitoring campaigns. 

Fig. 5. New SMS Vision camera prototype installed subsea. Image: 4Subsea.
Fig. 5. New SMS Vision camera prototype installed subsea. Image: 4Subsea.
About the Authors
Harald Holden
Harald Holden is Lead Engineer at 4Subsea within the area of drilling and wellhead systems. He holds a master’s degree in mechanical engineering from the Norwegian University of Science and Technology and has more than 20 years of experience in the oil and gas industry.
Related Articles
Connect with World Oil
Connect with World Oil, the upstream industry's most trusted source of forecast data, industry trends, and insights into operational and technological advances.