Automatic generation of structural geometric digital twins from point clouds

Area of Science:

• Digital Engineering
• Structural Health Monitoring
• Computer Vision

Background:

• Digitizing structural health monitoring (SHM) requires advanced digital twin (gDT) models that utilize 3D geometric data.
• Existing methods for SHM often lack efficient ways to integrate real-time structural data into digital models.

Purpose of the Study:

• To present a novel framework for generating and updating geometric digital twins (gDTs) of existing buildings.
• To infer semantic information from as-is point clouds for creating accurate virtual models.
• To enable automated structural health monitoring through integrated gDT and Building Information Models (BIMs).

Main Methods:

• Acquisition of 3D geometric data using regular laser scanning.
• Extraction of geometric information (position, section shape) from point clouds via supervised classification and domain knowledge.
• Inference of structural member function and section type.
• Automatic generation or updating of Building Information Models (BIMs) as the virtual model within the gDT.

Main Results:

• Successful generation of as-is geometric digital twins (gDTs) for building structural members.
• Demonstrated efficiency and precision in creating virtual models from real-world construction data.
• Extracted dimensional outputs from the gDT for effective structural health monitoring.

Conclusions:

• The proposed framework effectively leverages 3D geometric data to create and update digital twins for SHM.
• Integration of laser scan data with BIMs within a gDT enhances the digitization of structural health monitoring processes.
• The model shows significant potential for improving the accuracy and efficiency of monitoring existing building structures.