The imagery provided as a result of the data capture via drone will be used to produce a photogrammetry model to capture the rough 3D structure of the risers. By using our solution to analyse the high-resolution imagery, we will remove subjective manual review of the data by processing the images within a significantly reduced period of time.

The proposed general methodology involves capturing data at each Capture Position, which lies on a cylindrical grid. Abyss will define the application-specific details related to the AI System. This will enable us to align our image labeling standards so that they meet both the requirements of the AI training system. Abyss has a mature AI training suite which utilizes supervised deep learning models. We may use conventional machine vision techniques as preprocessing steps in training as needed. As we train the first machine learning models, we will be generating the deliverables using human-labeled data until the models are mature enough to utilize the AI alone.

In this system, Photogrammetry is used to:

  • Generate a 3D model of the risers.

  • Localize the cameras so that the defect and asset detections may be projected to the 3D model in the Projection Submodule.

  • By fusing with GPS to scale the model, make linear, area and volumetric measurements of the defects and assets.

The result is that the defects and asset detections are represented on a 3D point cloud, and each pixel can be denoted in spatial Cartesian coordinates in the same global reference frame. This forms a type of simple 3D digital twin, tagged with the asset types (flange, pipe, splashtron, etc.) and the defects (corrosion, dent, marine growth, etc.). All the information is present but not yet analyzed and compiled into a report.

The report is generated automatically, by rendering the 3D model at salient camera positions associated with the asset component and its position. There is one subsection in the report for each component in the asset, with the human-readable version of the feature vector displayed besides the rendering of the asset and the defects.

The value gained using drone services and AI resulted in a faster reduction of overall operational risk.

It is understood that there is a current backlog of inspections related to risers across offshore platforms. The time required to carry out inspections of all risers using traditional rope-access methods is typically 3-4 riser inspections per day. We believe that through the use of drone data capture, this can be considerably increased. This will unlock speed and cost scalability of this solution by enabling multiple drone teams to operate in parallel to result in a significantly faster completion of the inspection of risers.

Offshore inspection teams are integrating Abyss Solution technology in order to demonstrate automation and machine learning (ML) capabilities by providing insights on platform risers. Abyss Solutions will leverage inspection insights gathered during the first data capture campaign and develop machine learning models for automated condition assessment and reporting.

With the primary objectives above, Abyss aims to develop an automated reporting system that reduces the manual time required to create reports per asset using Abyss’ data processing tools. Abyss aims to reduce the manual effort by 50% by the end of the project, opening the door to scalable inspections across many assets with a quick turnaround time.

You can access this article if you purchase or spend a download.