16.11.2026 08:12

Making infrastructure inspections safer and more reliable

Švajčiarsko École polytechnique fédérale de Lausanne EPFL Autor neuvedený
EPFL professor Dimitrios Lignos and his team developed an automated, data-driven framework to help inspectors assess damage to steel infrastructure more consistently and safely. Combining 3D laser scans, computer simulations and AI models, it estimates whether damaged components can continue carrying loads and accounts for uncertainty in field measurements; the system is intended to guide, not replace, human inspectors. The researchers focused on electrical transmission towers and standardized steel profiles, and say the method could also be applied to bridges and other structures. They are seeking pilot projects and exploring drone-based inspections and ways to identify steel components for reuse.
AI zhrnutie

EPFL researchers have developed an automated inspection framework that helps engineers evaluate damage to critical steel infrastructure more reliably while reducing risks for inspectors in the field.

Telecommunication lines, electric transmission towers, steel bridges, and ski lifts are exposed to a wide range of natural hazards, such as strong winds or ice accretion as well as aging. Ensuring that these structures can operate safely depends on regular inspections by trained experts. These rely on visual assessments, which are then reported using specific protocols to detect deformations, corrosion, flaws, or cracks. However, reaching these sites can be challenging, conditions may be dangerous, and the consequences of a partial evaluation may be costly.

To provide a more evidence-based assessment, Dimitrios Lignos, professor at EPFL's Resilient Steel Structures Laboratory, and his team developed an automated data-driven inspection framework that complements human expertise. Published in Computers and Structures, the methodological framework combines scanning technology, computer simulations, and AI models to evaluate damage more consistently and assist the inspectors.

A guide more than a substitute

The inspection begins with detailed 3D scans of structural elements using laser technology, which capture their geometry and any visible deformation. The scanned data are transformed using point-cloud techniques and analyzed using algorithms trained on thousands of laboratory experiments complemented by virtual test samples validated through physics-based simulations. The system then assesses how much load a damaged element can still safely withstand and whether it can continue operating as intended or requires intervention. “The methodological developments do not attempt to replace human inspectors but to guide them to make more informed decisions,” says Lignos.

This framework is particularly valuable when inspectors face challenges that can delay inspections, prolong service interruptions, and increase maintenance costs. For this reason, the researchers focused on electrical transmission towers and infrastructure with standardized steel cross-sectional profiles. These are a critical part of Switzerland’s infrastructure, which are often located in remote areas and can be challenging to access after extreme weather events, such as snowstorms.

Dealing with real data

The methodological framework can extend to a wide range of structures, including conventional steel buildings and steel-concrete composite bridges, including railway infrastructure. In the same manner, this framework could also assist in determining whether decommissioned structures or components may be safely reused.

One of the framework's most distinctive features is its ability to account for uncertainty in data acquired in situ. Measurements collected through scanners are never perfect. Factors such as scan resolution and environmental conditions can introduce errors during the acquisition, known as measurement noise. This can influence how damage is interpreted and potentially affect the reliability of the predictions. The proposed methodology quantifies this uncertainty and evaluates its impact on the final assessment, allowing inspectors to better understand the confidence they can place in the results.

Further interdisciplinary collaboration

Researchers are now seeking pilot projects to take the technology from the laboratory into everyday infrastructure management. “The feedback obtained from inspection companies can provide essential information about how to improve and refine the established workflow before its eventual commercialization,” says Lignos.

Furthermore, Lignos and his team have been exploring alongside Alcherio Martinoli, professor at the Distributed Intelligent Systems and Algorithms Laboratory, the implementation of drone technology to reach structures in remote locations or that pose a risk for human inspectors.

Another ongoing project in collaboration with Olga Fink, associate professor from the Laboratory of Intelligent Maintenance and Operations Systems, aims to extend this framework to develop predictive algorithms for potential identification of members in steel structures to be reused after the end of their service life.



https://news.epfl.ch/news/making-infrastructure-inspections-safer-and-more-r
Autor
Spoločnosť / Organizácia
École polytechnique fédérale de Lausanne EPFL
Stiahnuť
Zdieľať
Galéria