Giving our bridge a 'voice' with IoT technology

The Sydney Harbour Bridge (SHB), like many iconic structures around the world, is ageing and requires regular preventive maintenance. Frequent inspections are essential and best practice but this alone may not be the most cost-effective and efficient way to maintain the majestic ‘coathanger’ in peak condition. What if the SHB could talk to us and tell us when she needed attention? Well, using the latest computer science and data techniques, Australia’s ICT Research Centre of Excellence - NICTA - is effectively doing just that.

The Structural Health Monitoring project being developed by NICTA for Roads and Maritime Services (RMS) uses lightweight, low-cost sensor technology and advanced data analysis to provide real-time monitoring of the structural health of the SHB and the safety of its concrete deck.

“In essence, the bridge tells us when it thinks it needs attention,” says Peter Runcie, business leader - structures NICTA. “We can sometimes think of this as giving our infrastructure assets a ‘voice’.”

What inspired the project?

“RMS wanted to extend the life of certain bridge components without significantly increasing maintenance expenditure,” said Runcie.

The bus lane on the eastern side of the SHB was built in the 1950s to replace the tram line that was there previously. Underlying this lane are almost 800 supporting components that bear the load of the 1.2 km-long roadway above. These concrete and steel components have been exposed to salt spray and weather over the years and are showing signs of their age.

Physical access is difficult. In order for RMS to conduct its regular two-yearly inspections on this part of the structure, workers need to either abseil down, use cherry pickers to come up from the ground or access via the maintenance gantry that travels under the bridge.

Runcie says what RMS required was an early-warning system that could monitor the condition of this part of the bridge between inspections and help determine when preventive action was needed. This would be complementary to the existing ongoing scheduled inspection program.

“When RMS could not find any technology on the market that could help them do what they wanted, they came to NICTA and we had to invent it.

"We like solving difficult but worthwhile problems, so it was a good fit for us,” says Runcie.

Why use IoT technology?

One of NICTA’s strengths is increasing productivity through data analytics and optimisation.

The use of Internet of Thing (IoT) technology in this and many other projects is due in part to the advancement of communications technology which has increased connectivity of people - ie, we all now have mobile phones and access to apps and data. It’s also due to the costs of hardware continuing to come down. This makes it is possible to increase the connectivity of things (ie, in this project it was sensors). Being able to use a greater number of sensors and a more capable communications networks provides the ability to access much more information than ever before.

However, Runcie says: “In order for IoT technology to be really useful, we need to pay a lot of attention to how the data is analysed and how information is then made available to users in meaningful and useful ways.”

How was the system designed?

The NICTA-developed system uses a combination of low-cost sensors (about 2400 required to monitor this part of the bridge) and several data analysis techniques, including machine learning, to analyse vibrations caused by traffic passing over the bridge.

Using innovative algorithms and simple components, the system can detect movements in the concrete deck that are considered to be abnormal, while ignoring normal movements due to vehicle traffic.

“Off-the-shelf equipment was too expensive then, and still is unaffordable,” said Runcie. “So we designed and built some hardware and a new sensing system using accelerometers (inexpensive accelerometers like you will find in a mobile phone). Each accelerometer is controlled by a small computer, also low-cost technology, which just a few years ago would not have been feasible.

“All the sensors are connected by weatherproof ethernet and a fibre-optic network. Some data is processed by the computers on site and the results of that processing together with some of the raw sensor data is sent back to NICTA’s data centre in Canberra.”

Data analysis is the key that unlocks the value in all the data that is collected. NICTA’s machine learning research group has recently been recognised as being in the top five in the world. This team of scientists has developed new analytical techniques that have been applied here.

“The neat thing about machine learning analysis is that it is a data-driven approach that can be applied even when the physical structure or system is very difficult or not feasible to model in an engineering sense,” says Runcie.

If any anomalous readings are detected, the asset manager and bridge inspectors are notified by email and text message so they can schedule an inspection. They also have a web-based application to monitor the bridge - this is a decision support tool that helps bridge management schedule maintenance resources.

“What RMS wanted us to do was to interpret the information for them and just give them the answer that they wanted,” says Runcie. In other words they wanted to know: Do I need to go schedule an inspection or don’t I?

The system is therefore designed to be easy to use for the asset manager. Basically, it provides a web page that shows the map of the bridge which is colour coded. “If it’s all green, everything is good; if orange, it means that some anomaly has been detected and an inspection should be scheduled,” says Runcie.

“They still do their two-yearly inspection but this just gives them additional situational awareness in a form they can understand.”

Can this technology be applied to other applications?

This technology is certainly applicable to other bridges and structures around Australia and internationally says Runcie. “The data that is needed for each application is going to be different. For example, it may not be the vibration around the joint that is important, it could be stresses and strains, fatigue cycles or corrosion that’s important, or some combination of these. It could even include other information such as maintenance and inspection records, loading or environmental information. The analytical techniques are then adapted for each application.

“We have made the system design generic in terms of the sensors and data sources that can be used and also the analytical techniques that can be applied. If new analytical algorithms are needed they can be developed and easily implemented as software in the system,” he says.

Although it is a bit of a conceptual mind shift for engineers and some asset owners, Runcie points out that data-centric analysis techniques such as this are complementary to more traditional approaches and provide new insights into the condition and performance of structures