T2 judgement day

This AI solution detects, classifies and monitors the deterioration of building envelopes and infrastructure.

Late in 2020, Thornton Tomasetti announced the launch of T2D2, a software-as-a-service (SaaS) platform that uses artificial intelligence (AI) to identify and assess damage and deterioration to building envelopes and structures through images. It applies data from more than 50 years of building inspection and forensic investigation work to identify damage and defects before they become major issues.

CEO Badri Hiriyur PhD, Director of CORE AI at Thornton Tomasetti, said the AI within T2D2 has been applied to identify visual anomalies and other signatures of deterioration on façades and structures. It analyses digital images or videos taken by any type of camera (hand-held, fixed or drone-mounted) and can detect and identify conditions of deterioration. “The emergence of big tech with big computing and big data made it possible. T2D2 uses the same deep convolutional neural networks powering state-of-the-art computer vision models. It is trained using a cluster of Nvidia GPUs [graphics processing units] on a vast dataset built from tens of thousands of images from Thornton Tomasetti’s own database of building structures and façades as well as bridges and other infrastructure.

“Using data acquisition and analysis and telemetry to replace traditional methods of inspection saves time and money – and there can be tremendous value with its continuous use. It will identify even the smallest anomaly, such as a hairline crack, which might not be picked up between prescribed inspection cycles but can deteriorate quickly and lead to costly problems.”

The AI programme knows when it sees a crack or a change in previously detected damage. The detected damage conditions can be geolocated on structures using CAD, BIM or photogrammetric 3D models and presented in digital assessment reports that get updated in real time. The system also continues to improve and refine its findings over time, as it ‘learns’ the specifics of each asset/structure. While data acquisition in the form of images is key – particularly the use of drone footage for inaccessible and high areas of the building – it is the AI’s knowledge and judgement that are vital. “The disruptive innovation here is in the extraordinary training we have given T2D2 to analyse images and recognise deterioration,” notes Hiriyur.

Tried and tested

The technology has been tested on numerous building envelopes and structures across the US and Canada. It has been used to scan for damage and deterioration on roofs of million-square-foot industrial properties in California, joints on historic stone bridges in Maine, brick and stone façades in New York City and Chicago, stucco façades of residential towers in Florida, concrete transportation systems, water treatment facilities in New York and more.

As a testament to its abilities, T2D2 has also been selected by tech giant Nvidia to join its Inception programme, a virtual accelerator directed at AI start-ups in various industries. It was also named a winner in the New York City’s Department of Buildings’ inaugural Hack the Building Code innovation challenge that called for ideas on how to improve the city’s buildings, keep construction workers and the public safe, and modernise the construction process.

“It is more than an inspection tool,” said Thomas Scarangello, Executive Chair of Thornton Tomasetti. “It will help our clients better manage their structures and building envelopes, making them more resilient, safer and more economical to maintain throughout their full life-cycle.”    

For more information, see t2d2.ai