• La Société Wallonne des Eaux

    Deep Convolutional Neural Network on 3D Reality Mesh for Water Tank Crack Detection

    Juprelle, Liège, Belgium

Project Summary

    • REALITY MODELING - La Société Wallonne des Eaux

Determining How to Repair Water Tower Damage

Société Wallonne des Eaux (SWDE), a regional water corporation, owns and maintains a series of water towers throughout Belgium. One of these towers, built in 1981 in Juprelle, features an exterior with siding brick, a concrete-filled slide, and a terracotta supporting wall. Though filling the slide with concrete reduced the stresses on the brick and provided a strong link to the inner structure, it also caused condensation to form on the interior walls, resulting in significant degradation. The damage included burst joints, the separation of edifice bricks, and multiple cracks. SWDE needed to quickly conduct a survey to determine how to repair the tower and ensure continued water service to the area, but using elevators to lift workers up to the tank or taking photographs from the ground was inefficient and often did not reveal the full extent of water tower damage.

Safer Surveying with Drones

SWDE determined that they should use drones to conduct a full survey of the damage. Not only could drones provide a better assessment of conditions than traditional surveys, but they would also allow workers to perform the task from the safety and comfort of their office. However, even the improved vantage points provided by drones still required human interpretation of the current conditions, which ran the risk of overlooking small cracks. SWDE wanted to go beyond relying on raw imagery and find a way to automate crack detection, which would further improve the surveying process and ensure that they could repair all damage in a single session.

Automatically Detecting Tiny Cracks with Artificial Intelligence

During discussions with representatives of Bentley Systems, the SWDE repair team learned that there were artificial intelligence capabilities for ContextCapture in development that can automatically detect cracks in infrastructure. SWDE and Bentley worked together to refine the module and test it out on the damaged water tower. They first captured 3,000 images from drone surveys, then used them to create a reality mesh of the tower within ContextCapture. Once the reality mesh was complete, the application used AI to scan it and detected 1,704 cracks in the tower, with 520 of them having a crack length of less than two millimeters. The inspection also detected that the façade had detached more than 10 centimeters from its original placement, which would have been overlooked during traditional assessments.

Improving Accuracy while Shortening Survey Time

As a result of the pilot project, SWDE formed a specialized drone and scanner survey team, which uses reality meshes and artificial intelligence for crack detection. In addition to detecting more cracks with smaller lengths than previous methods, the new method reduces the time needed for surveys by as much as 66% and helps them accurately plan what renovation work is needed, saving significant costs. They also plan to conduct additional surveys of assets after they are repaired, so they can establish a baseline of conditions they can compare to future surveys undertaken in five to 10 years. The digital replicas also help SWDE determine how to prevent degradation and lower the risk of any interruptions to the water supply.

Project Playbook: ContextCapture, MicroStation, Pointools

  • Société Wallonne des Eaux needed to quickly survey the damage to a water tower so they could begin repairs and ensure continued water service.
  • The repair team worked with Bentley to develop artificial intelligence capabilities for ContextCapture that could use drone imagery to detect cracks.
  • AI-enhanced drone surveys reduce the time needed by as much as 66% and helps SWDE accurately plan what renovation work is needed, saving significant costs.
  • The digital replicas also help SWDE determine how to prevent water tower degradation and lower the risk of any interruptions to the water supply.
  • ContextCapture Insights allowed us to quickly model and identify the structural defects in the water tower and budget accurately for the required renovation work. Without the use of the drone and Bentley software, we would have missed major structural issues.”

    Christophe Taelman Engineer and Deputy Director of the Design Office La Société Wallonne des Eaux