Saad Hamid, Production Engineer at Saudi Aramco
Abdullah AlDraihem, Technical Advisor at EV Offshore KSA

In a recent interview with Electrek, Elon Musk said:

“When radar and vision disagree, which one do you believe? Vision has much more precision, so better to double down on vision than do sensor fusion     …. As vision processing gets better, it just leaves radar far behind.”1     

Advanced imaging technology has become an indispensable part of the modern world. It is prevalent in the latest generation of automotive vehicles. Tesla, among several      car manufacturers, uses computer vision and advanced image processing techniques to support safer driving. In industries involving high levels of security, advanced imaging plays a critical role in delivering the data and accuracy required for critical decision-making processes, such as facial recognition. The quality and precision of the image data acquired are decisive factors for these applications.

Figure 1: Tesla Computer Vision (Courtesy Electrek.co)

The oilfield is no stranger to imaging and cameras either. Downhole video cameras have operated in      oil and gas wells since the 1980’s, but it is only in recent years that the technology has advanced sufficiently to reliably capture, store and process the large volumes of high quality      images and high frame-rate      videos. This information is required for computational analysis and      as input for machine learning models for deriving meaningful outcomes through Artificial Intelligence      (AI).

Advances in computer vision techniques and image processing have led to the ability to make detailed and accurate measurements from images – known as visual analytics. This new data analytics source, combined with state-of-the-art acquisition technology has significantly improved the understanding of complex well issues while reducing operational time, risk, and cost. Examples of its applications include hydraulic fracturing, well integrity, erosion, restrictions, and leaks.

To exploit the information value from what is now a suitably rich visual dataset, bespoke software applications have been developed to analyz     e and quantify the captured video data. With this software system, the user can measure the dimensions of captured objects within the field of vision of the camera system by placing markers, shapes or freeform drawings directly onto the image. The system defines these sub-pixels dimensions highlighted by the operator and converts to calibrated real-world measurements of distance, thereby returning quantification of lengths, areas, arcs and azimuth in any plane in all three dimensions.

Figure 2: Optis Infinity Multi-array Sideview Camera (Courtesy EV)

Multi-array side view camera system

The developments in video acquisition systems and computer vision technology discussed above provide a step-change in the value and number of applications for downhole video. This latest camera system incorporates the innovative arrangement of four axially aligned image sensors contained within a novel and      unique lens system. The system operates by capturing all four side-viewing camera feeds simultaneously at 25 frames per second. A hybrid system of real-time and memory functionality is now possible through the integration of high-capacity memory modules and telemetry systems. This unique technology allows all four array cameras to record simultaneously and save the recordings memory storage while delivering high-quality live video at the surface. This quantum shift in information offers an unparalleled understanding to support real-time decision-making and enable effective well management.     

Image Stitching Process

To extend the application of      array technology, significant advances have been made in image processing to deliver a single, continuous 360-degree image derived      from the integration of images from each of the four individual cameras. This process, known as image stitching or ‘     mosaicking’. Stitching      takes full advantage of the overlap in the circumferential direction between adjacent cameras and high frame rate video captured by each camera. These two features in combination ensure that features of interest are captured by at least one camera and over      successive depth/time points, ensuring that recorded datasets capture a complete & quantifiable image of the wellbore.

An example of the input video and the resulting output images from this image stitching process is shown below in Figure 3 for a perforated area of a well. With current array camera technologies this process analyzes in excess of 5×109 input pixels (> 5 Gigapixels) to produce an image of 10m (approximately 33ft), with resulting file sizes > 20 MB/     m ( > 6.5 MB/ft). The resulting images quickly becomes very large in size,      the mosaic images are therefore stored in short, depth indexed sections as      the database and accessed by the playback software in sequence to produce images over any required interval      with the minimal      image processing and file handling.

Figure 3: Example mosaic image over perforated area of well based on input sequence      (Courtesy      EV)

Another advantage of the image stitching process is that the technique is purely driven by the properties of the images and does not rely on external sources of information regarding tool speed or the centralization/offset of the tool within the wellbore. The benefit of this is that the algorithm is able to compensate for changes in downhole speed, centralization and tool orientation, providing a measure of each effect as an output. 

This in itself provides significant opportunities for downhole depth and speed correction which may assist greatly in improving the location/deployment of mechanical services, such as cutters and perforating systems, deployed in combination with the array video technology.      Similarly,      time-lapse analysis of well integrity or cased hole petrophysical logs may be significantly improved by enabling more accurate depth correlation and motion compensation of successive datasets.

Log data integration and 3D viewing

Having successfully created seamless, continuous images of the wellbore environment, it is now possible to integrate the data with other sources of data           on a time or depth-matched basis. By integrating      this well or structural integrity logs, it is now possible to visualize the combined data in both 2-dimension and 3-dimensions through the development of a combined video/log data analysis suite, as illustrated in figure 4. By viewing and analysing multiple datasets with the      mosaicked image, it is possible to gain a greater understanding of the well condition and behaviour as well as quantifying the changes that may have occurred. Through this technique, it is also possible to identify root cause and assess the impact and severity of the issue for improved decision making with regards to remedial actions with the added benefit of simplified and      visually intuitive information.

Figure 4: Simultaneous 2D presentation of quantitative multi-finger calliper data and qualitative 360-degree mosaic image (Courtesy      EV)

     For further enhancement, the mosaic images can be superimposed onto 3-Dimensional CAD models and technical drawings of components and assemblies as supplied by the original manufacturer. This allows the image to be seen in its intended geometry, and from any viewing angle or perspective, enabling identification of variation between original      specifications and ‘as measured’ geometries.

Case Study

The Challenge

The following example comes from an operation in Saudi Arabia, where the well in question was completed as a single lateral, horizontal gas well, with a 5 ½” pre-drilled liner in 8 3/8” open hole. Workover operations were completed previously to replace the corroded carbon steel with 4 ½” S13% Cr tubing. 

During a drift run, a restriction was encountered near the upper R-Nipple at 2,425 feet. A 2.75in lead impression block was run, however the investigative efforts were inconclusive. With the well shut in, the operator needed to characterize and quantify any identified restrictions in the production tubing. 

The Solution

A high-resolution camera combined with a multi-finger      calliper string was deployed on E-Line. The combination of both visual and quantitative data delivers a more complete answer to well integrity challenges. This leads to enhanced interpretation and provides invaluable 360° pipe coverage to compliment the limited radial coverage available from a standalone mechanical      calliper. Deployment on electric line means that multi-finger      calliper measurements and high definition down-view and side-view images can be acquired real-time, at surface.

The Results

High resolution video was logged from surface to the depth of interest, where the live footage instantly revealed parted tubing below the pup joints.  The camera identified the lower portion of tubing had dropped about 18 feet. The side view footage revealed the extent of the damage on the upper parted tubing in more detail (Fig.5)

Figure 5: Parted tubing identified by downview camera. Extent of damage revealed by side-view images. (Courtesy      Aramco/EV)

A 360 degree “     m     osaic     ” image was generated from the side-view footage to help visualise the extent of the damage in more detail (Fig. 6)     .

Figure 6: 360 degree ‘mosaic’ image of parted tubing damage. (Courtesy      Aramco/EV)

The multi-finger      calliper revealed strong evidence of galvanic corrosion and severe pitting within both pup joints (Fig. 7). The associated f     low c     ouplings showed little sign of corrosive damage, suggesting concentrated galvanic corrosion in the pup joints associated with mixed metallurgy. 

Figure 7:      Calliper pass over pup joints revealing severe corrosion and confirming parted tubing damage.      (Courtesy      Aramco/EV)

The combination of qualitative and quantitative data provided a holistic view of the situation downhole and allowed the operator to move ahead safely with a workover operation. 

Conclusions

Modern downhole video provides high-definition images, and high frame rates to identify and diagnose wellbore features that are otherwise unobtainable by other technologies. The quantification of these images, and the ability to combine with other quantitative log datasets, provide      a more intuitive visualization of complex information and greater accuracy of interpretation.

The continuous 360-degree images of wellbore conditions, enabled by advances in computer vision, image processing techniques, and downhole video technologies, deliver a level of wellbore evaluation that far exceeds the limits of conventional logging technologies. The application of visual analytic techniques to the images from these latest generation technologies has enabled the development of new diagnostic methods. These methods provide an improvement in information provided      which empowers decision-making, leading to      better economics and reduced operational risks.

Leave a comment

Your email address will not be published. Required fields are marked *