Civil Engineering: Monitoring bridge health using smartphones

Crowdsourced data from the smartphones of vehicle passengers crossing bridges, such as the Golden Gate Bridge in San Francisco, may help monitor bridge structural integrity, reports a Communications Engineering paper. The findings suggest that crowdsourced monitoring could be a cheap and convenient way to monitor the structural integrity of transportation infrastructure worldwide, and could potentially increase the lifespan of bridges by up to 30%.

There is a global need for infrastructure monitoring to improve the resilience and longevity of bridges, buildings, and other structures. The structural health of bridges is usually visually assessed by engineers on-site, which is often time consuming and infrequent, or measured using static sensors incorporated into the bridge, which are expensive. Measuring the vibrational frequencies of bridges has previously been used to identify bridge damage and deterioration, but the data to support this approach have been limited.

Thomas Matarazzo and colleagues used anonymised crowdsourced data from mobile phone accelerometers (devices that detect vibrations) while running the Uber app during everyday trips over the Golden Gate Bridge, a long-span suspension bridge, to capture its vibrational frequencies. The researchers found that these frequencies could be identified to within 3% accuracy using less than 100 datasets from different trips with different phones. Further studies also demonstrated the capability of the approach to monitor a more commonly found short-span highway bridge in Italy. The authors suggest that data from crowdsourced monitoring could increase the lifespan of an older bridge by about 2 years (15% increase), and add about fifteen years of service to a brand new bridge (30% increase).

Although the approach has not yet directly identified structural degradation, the findings highlight the potential of using crowdsourced data for cheap and continuous monitoring of infrastructure, irrespective of vehicle model, speed, or phone model. By expanding to very large volumes of crowdsourced data, the approach could harness advances in artificial intelligence and computational learning for further improved sensitivity and monitoring capability, the authors suggest.