• Get Started
  • Documentation
  • Samples
  • Blog
  • Pricing
    • Free Trial

    Machine learning in iTwins

    Why use Machine learning in iTwins?

    Machine learning (ML) brings new possibilities to infrastructure projects by automating data analysis and uncovering patterns in complex datasets. Within the iTwin Platform, ML helps streamline tasks like analyzing reality data, extracting information from legacy diagrams, and integrating advanced models for predictive insights. This use case highlights some practical applications of ML in iTwins.

    Key scenarios

    Intelligent analysis of Reality Data

    Reality data—such as photos, point clouds, and meshes—often requires significant manual effort to analyze. The Reality Analysis API applies machine learning to automate this process, making it easier to identify anomalies, detect defects, and classify assets and features. By automating these tasks, teams can quickly spot issues or gain valuable context, improving decision-making and project efficiency.

    Example 1 – Feature detection to detect components inside a tunnel point cloud. Courtesy of Yunex.
    Example 1 – Feature detection to detect components inside a tunnel point cloud. Courtesy of Yunex.

    Example 2 – Asset classification for identifying traffic signs on a road. Courtesy of Slovenska.
    Example 2 – Asset classification for identifying traffic signs on a road. Courtesy of Slovenska.

    Automating PnID data extraction

    Legacy Piping and Instrumentation Diagrams (PnIDs) are often stored as raster images, making data extraction difficult and time-consuming. The PnID to iTwin API uses machine learning to automate this process by identifying key components like valves, pumps, and vessels. It also applies Optical Character Recognition (OCR) to extract both typed and handwritten text labels, which are then used to tag elements in the model for easy navigation. This converts legacy data into a structured, digital format that seamlessly integrates into the iTwin.

    Example 3 – Element classification in a PnID. The extracted components are listed on the right.
    Example 3 – Element classification in a PnID. The extracted components are listed on the right.

    Integrating external ML models

    iTwins can integrate with custom or third-party machine learning models to address specific project needs. Using iTwin.js, teams can create custom applications that connect these models to iTwin data for tasks like predictive insights or risk assessment. The resulting insights are then visualized directly within the iTwin, helping teams make informed, data-driven decisions.

    Example 4 – An iTwin of a windfarm integrating an ML algorithm that predicts future energy production based on weather conditions.
    Example 4 – An iTwin of a windfarm integrating an ML algorithm that predicts future energy production based on weather conditions.

    Summary

    These are just a few examples of how machine learning can transform infrastructure projects within iTwins. From analyzing reality data to extracting insights from legacy diagrams and integrating custom models, ML opens up a range of possibilities for teams to explore. Looking ahead, as ML continues to evolve, its integration with iTwins will only deepen, providing new opportunities to refine workflows, improve decision-making, and stay ahead of emerging challenges.

    Featured APIs

    Reality Analysis API
    Use machine learning to classify reality data, detect objects and features in 2D and 3D for defect analysis, asset management, and more.
    PnID to iTwin API
    Transform PnID images into a functional model by classifying components and adding tag information for easy navigation.

    Featured Tutorials

    Running the PnID ML Pipeline
    Walk-through on how to run the PnID ML Model and get your results.
    Get started with the Reality Analysis API
    Understand the basic workflows of the Reality Analysis API.

    Was this page helpful?