To check the structure of a road bridge, the appearance of any faults, the behavior of the piles or deck over time, the structure can be surrounded with 240 sensors… or use two smartphones. This is the conclusion of an experiment by the Senseable City Lab of the Massachusetts Institute of Technology (United States), carried out on the Golden Gate Bridge, in San Francisco, and on a highway bridge in the southeast of the Rome metropolitan area ( Italy). The research article was published in early November 2022 in Communications engineering.
The team shows that it is possible to replace fixed sensors intended to provide data on the physical properties of a bridge (torsion, displacement, misalignment, etc.), with a view to maintenance and prevention, with classic smartphones built into the car, thanks to sensors such as the accelerometer, gyroscope, GPS, which such a device already contains.
Predictive maintenance by drivers
Vibration data is recorded as the vehicle passes over the bridge. And this new method proves to be just as reliable as the traditional approach. But much less expensive. The researchers go even further: if we multiply the data collection via smartphone by multiplying the average users, predictive maintenance will only become more efficient.
Specifically, the researchers made 102 trips on the Golden Gate Bridge, equipped with an iPhone 5 and an iPhone 6 placed on the foredeck, with the Sensor Play application downloaded and activated. They then checked the data collection. But they also recovered from routes they had no control over, namely 72 rides by Uber drivers. This corresponds to 37 vehicles and 19 different smartphone models. Furthermore, these drivers did not always respect the speed limit and in any case they did not have the same way of driving as the researchers.
But the latter considered that the famous Californian suspension bridge, 1280 meters long, was too atypical a structure to conclude from the beginning that the method of capturing the vibrations could be transposed. Hence the choice to experience it in Ciampino, Italy, on a 28-metre concrete bridge over an E80 junction.
A low error rate
This time the vibrations were recorded using an Android application developed by the researchers for road maintenance workers. It was the latter who made the 250 trips on the bridge from which the vibration data is taken; the harvesting conditions were therefore only partially controlled.
In the end, the difference between the values measured by the smartphones and those coming from the fixed sensors on the bridges is mostly small and makes the researchers’ technique very reliable. As for the Golden Gate Bridge, the team even estimates that between 10 and 50 smartphone datasets are enough to get results with an error rate of only 10%. “Once the error rate drops below 10%, writes in his article, adding 10 datasets tends to reduce this rate by 1%.”
The scientists also insist that a single car ride over a bridge is not enough to have useful measurements, but note that only a hundred data points from the passage of ordinary road users, and not from researchers applying a precise protocol, produce estimates with an error rate of less than 6%.