The Constructor

Machine Learning Software for Facade Inspection | Video Inside

Machine Learning in construction

Machine Learning in construction

Reading time: 1 minute

Machine learning and artificial intelligence have long been in the construction industry to make some real impactful ripples. After almost two years of development in Thornton Tomasetti’s CORE lab, engineers have created a computer-vision, machine-learning algorithm known as the Thornton Tomasetti Damage Detector (T2D2), that can help identify damage to a building’s exteriors through videos or images. 

The T2D2 software as a service (SaaS) platform is intended to speedily reveal any hidden aberrations that may go unnoticed during manual façade inspection, which is a tedious and challenging process.

The system will be somewhat guided, and the engineers will review the detections made by T2D2, which will be flexible enough to look for conditions across different material and structure types. It could be used for all types of structures, but the main goal is to look at aging structures that require periodic inspections.

To get an idea of the effectiveness of the tool, the team went through hundreds of drone images collected over decades of building inspections. The results were very promising and were capable of assisting human engineers before they get close to any building for inspection.

On a typical image, T2D2 runs several steps. Firstly, using its cache of knowledge acquired through thousands of previously annotated images, it quickly spots the geometry of the structure and assigns material types. It then does a pass to spot potential damage that a material type is vulnerable to. Finally, it produces an emboldened version of the picture, with potential damage classified and highlighted for review.
Inspection of cracks using T2D2

T2D2 was initially trained essentially for concrete structures, but now the capabilities have been expanded. The tool can now easily identify and classify damage on masonry, brick, stucco, and other commonly used materials.

Due to the inclusion of machine-learning elements, the algorithm improves every time it is put into application. Further, the engineers plan to use reinforcement learning to improve the models. With that, an engineer would be able to mark the false positives and negatives, helping to enhance the results.

With T2D2, the way of inspecting buildings is being redefined. Thornton Tomasetti engineers have been working in partnership with drone survey companies to provide a thorough façade inspection service.

Building owners usually perform façade inspections just to meet regulatory demands, but the comparatively lower expense of using T2D2 accompanied with the use of drones may change that.

Frequent inspections would mean building owners can make small fixes to prevent bigger problems while saving millions in repairs.

Read More

Self-Assembling Robot Swarms to Form Bridges and Buildings: The Future of Construction

Boston Dynamics’ Robot dog Supervises Construction Site in London | Video Inside

A New Drone-based System to Assess Aging Infrastructure Projects

Exit mobile version