Assets such as buildings and bridges are typically maintained when something goes wrong or according to preventative maintenance schedules. This is risky and inefficient as maintenance is either too early or too late. Structural Health Monitoring (SHM) is a condition-based technology to monitor infrastructure using sensing systems.Data61 has been working with Transport for New South Wales to develop SHM systems to monitor the structural conditions of the iconic Sydney Harbour Bridge and other small bridges in Sydney. We have also developed machine learning techniques to detect road pavement potholes and monitor machinery assets such as building HVAC systems and ship engines.
Damage Detection in Civil Structures
Data61 has patented a machine learning technique that identify damage on civil infrastructures like bridges using only data from healthy conditions of the structures. It is described in the framework in the figure below. The technique fuses healthy vibration data collected by sensors (e.g. accelerometers) instrumented in the bridge to model the bridge normal behaviour. By looking at how different newly collected data are from the normality model, anomalies or damage can be identified.
The technique has been used to monitor the Sydney Harbour Bridge. In this case study, the bridge asset manager required a way to remotely monitor each of the 800 structural jack-arch supports under the bridge deck. The objective is to provide early warning of any problem so that preventative maintenance can be carried out without disrupting road users. In addition, a business goal was to extend the life of the asset without a significant increase in maintenance expenditure. Current practise involves a visual inspection once every two years and some jack arches are very difficult to access.
Data61 have developed and implemented a large scale SHM system that uses 2,400 sensors to continuously monitor each of these 800 structural components. The system continuously monitors the sensor data, applies machine learning and other data analysis to detect anomalous behaviour. A web-based application (see figure below) is provided that shows current bridge condition and specific bridge components warranting closer inspection. When early signs of a problem are detected, email and text message alerts are sent to the asset manager and bridge inspector so that an inspection can be scheduled. This SHM system was awarded “The Most Practical SHM Solutions for Civil / Mechanical Systems Award” at IWSHM - the 10th International Workshop on SHM in 2015 (a top conference in SHM).
Besides we also instrumented and monitored the two other bridges in Sydney, including a cable-stayed bridge as shown below.
The technique is generic and can be extended to detect anomalies in other assets or application domains. We have used a similar technique for road condition assessment using sensors mounted in vehicles. The technique was successfully validated in a trial using school buses in Wollongong. Moreover, we have also worked on fault detection in machinery assets such as building HVAC systems and ship engines.