Project image

Business Goals

  • Detect the corrosion, identify the surface covered with rust, and its severity.
  • Automate inspection of oil platforms.

Challenge

  • Marine and offshore structures maintenance are costly due to the involvement of expensive personnel and the need to deliver them to the facility. Besides, the losses include stopping the oil production for the inspection period. The venue was supposed to produce nearly 130 thousand barrels of crude oil during this time!
  • Some surfaces are physically unavailable for a human inspection, for example, pipes over the ocean or hard-to-reach places inside the platform.

Results

  • The solution demonstrated 100% accuracy in near real-time rust detection (classification into `contains rust` and `doesn’t contain rust` categories) and 0.67 Intersection-over-Union (IoU) for segmentation of corroded regions.
  • The solution was integrated as a module of an oil platform digital twin, allowing real-time facility assessment and access to hard-to-reach locations.

Implementation Details

  • First, the data strategy was designed, and appropriate data were collected and labeled.
  • Then, the images were classified into containing rust and not containing rust.
  • Lastly, a number of state-of-the-art segmentation models (UNet, FCN, PSPNet, etc.) with various backbones (ResNet, Inception, EfficientNet, etc.) were trained, validated, and the best model was chosen.

Industry

Keywords

  • Oil & Gas
  • Computer Vision
  • Segmentation
Roadmap
Solution Design
AI Solutions Architect
Data Collection
AI Solutions Architect, Computer Vision Engineer
Data Labeling
Labeling Team
Exploratory Data Analysis
Computer Vision Engineer
Data Preprocessing
Computer Vision Engineer
Model Development
Computer Vision Engineer
Model Performance Evaluation
Computer Vision Engineer
Hyperparameters Tuning
Computer Vision Engineer
AI Service Design
AI Solutions Architect, MLOps
AI Service Coding
MLOps
Release

Sign up to receive the project description

    Roadmap
    Solution Design
    AI Solutions Architect
    Data Collection
    AI Solutions Architect, Computer Vision Engineer
    Data Labeling
    Labeling Team
    Exploratory Data Analysis
    Computer Vision Engineer
    Data Preprocessing
    Computer Vision Engineer
    Model Development
    Computer Vision Engineer
    Model Performance Evaluation
    Computer Vision Engineer
    Hyperparameters Tuning
    Computer Vision Engineer
    AI Service Design
    AI Solutions Architect, MLOps
    AI Service Coding
    MLOps
    Release

    Relevant Projects