Wireless monitoring extends lifetime of critical infrastructure
Wireless sensing networks are being used to improve the safety of critical infrastructure such as bridges and even helicopter blades.
When the Minnesota Bridge over the Mississippi river collapsed during an evening rush hour in August 2007, killing 13 people and injuring 145, the parlous state of America's civil infrastructure became headline news. Far from this being a rare event, apparently more than 130 American bridges collapsed between 1989 and 2000: that's about one a month.
Europe's bridges seem more robust than those of the US, thanks to a more rigorous approach to inspection. Nevertheless, in August 2009 commuters using the Belfast-Dublin rail line had a lucky escape when part of a viaduct fell into the sea just after a train carrying 50 passengers had crossed it. Thankfully the train driver spotted subsidence and raised the alarm to halt any other trains on the line.
Such catastrophic failures are boosting interest in embedding permanent sensors and communications links into all kinds of infrastructure to monitor it for structural deterioration. While safety is the main incentive, there is growing interest in how these techniques could extend the life of structures - on the basis that the 'greenest' infrastructure is likely to be the infrastructure you repair rather than replace.
"If you paint your house every five years, the wood won't rot. If you leave it for 15 years, you might end up having to replace all the windows," explains Mike Robinson from MicroStrain, an American company that specialises in wireless sensing systems for civil engineering, medical and military applications. "By doing the small jobs on a regular basis as they're needed - which you can determine with the structural health monitoring system - you don't have to do the major repairs later on."
Manual non-destructive tests, such as scanning metal plates with ultrasonic waves, have been used for decades in bridge maintenance in Europe, and in the nuclear and civil aviation industries. Permanently installing the piezoelectric transducers, accelerometers, and strain gauges that are needed to measure these effects is a different proposition. Do you place the sensors under known weak spots, or everywhere? How do you power them? How long will they last? How do they communicate? How do you
interpret their measurements - do you use them to build a computer model, to generate an image, or do you just take single readings?
Energy harvesting for health sensors
The big idea in structural-health monitoring systems is that they should be self-powering networks that can collect and communicate data wirelessly for decades. Energy harvesting is becoming feasible, as MicroStrain has demonstrated in a number of projects. One involved sticking piezoelectric materials directly to the pitch link of a helicopter (the part responsible for controlling the rotors' angle of attack) to convert its vibrations into stored electrical energy for operating wireless strain sensors. In a shipboard application, the firm has applied similar materials to harvest vibration energy from machinery.
MicroStrain has also deployed solar energy harvesters on the Goldstar Bridge in Connecticut, US, and the Corinth Bridge in Greece. The latter installation uses arrays of wireless tri-axial accelerometer nodes to monitor the span's background vibration levels to detect seismic activity, while the Goldstar set-up monitors not only vibration but also strains, as part of a long-term project to learn how bridge monitoring systems can be used to evaluate in-service behaviour.
The main power burden on these systems is transmitting the data they gather, so using lower-power radios and intelligently controlling how and when they transmit data can ease the burden on the energy-harvesting elements. MicroStrain is using Texas Instruments' CC2420 IEEE 802.15.4 wireless transceivers operating in the 2.4GHz ISM band but hopes to reduce energy consumption by a factor of 200 by upgrading to an ultra-wideband 802.15.4a radio made by DecaWave, a Dublin-based fabless semiconductor company. In addition, MicroStrain has smart comparators in all its wireless nodes that only allow current to be drawn if there is sufficient energy to complete the task in hand. "It's rather like balancing an energy chequebook," says Robinson.
The Laboratory for Intelligent Structural Technology (LIST), led by Jerome Lynch at the University of Michigan, is trying to reduce power consumption by doing more processing at the wireless sensor nodes, so that only a small amount of data is ever broadcast - an advantage being that the data takes up less space in the crowded unlicensed ISM bands. The idea is to use network-wide algorithms, in which spatially distributed wireless sensors exchange small amounts of data, to compute a complete picture of the patterns of motion of a structure at its resonant frequencies - also known as its modal properties. So, instead of updating a finite-element model that describes the physics of a system on an off-line computer, the computing resources of the wireless sensor network nodes share out the work of doing this.
"The idea is that the network operates in an ad hoc fashion, self-configuring on the fly to achieve certain computing objectives such as speed of computing, minimal power use, minimal bandwidth usage. We arbitrate those competing demands using the laws of free-market economy, just like a market full of buyers and sellers," says Lynch.
Working with other research groups, the Michigan lab has been involved in various trials of battery-powered wireless monitoring systems - for example at the Alamosa Canyon Bridge in New Mexico, US, which is kitted out with seven MEMS accelerometers, and the Guemdang Bridge in South Korea, which has 14.
Greener wind turbines
Its most recent paper, though, describes a collaborative project with the University of Hanover in Germany, furnishing wind turbines with MEMS accelerometers and metal-foil strain sensors. The results are being used to compute operational deflection shapes.
Undetected turbine damage can cause catastrophic failures, as happened to a turbine in Dunbar, Scotland in 2005, which resulted in £1.25m repair costs. Other reported damage mechanisms in wind turbines include corrosion, foundation failures, and fatigue cracking.
"By collecting the response data from sensors on the turbine, we can use model-updating to calibrate a detailed finite-element model of the turbine that can then be used for structural health monitoring purposes, " explains Lynch. The information could be used for damage detection and also to provide data to inform the design of more cost-effective turbines.
It's early days for this project. More work is needed on long-term monitoring using permanent systems to study the loads on - and structural responses of - turbines over a large range of conditions, in particular for offshore turbines battered by waves. For damage detection, algorithms would need to be developed and built into the computing cores of the wireless sensors to provide automated monitoring of the condition of turbine towers, nacelles, and blades in order to report developing and final failures. If this can be achieved, the cost and risk inherent in building turbines might be reduced, increasing their attractiveness to the energy industry and to the public as a whole.
Back on this side of the Atlantic, Dr Paul Wilcox and colleagues at the department of mechanical engineering at the University of Bristol, and a team at Imperial College London led by Professor Peter Cawley have been working on low-frequency acoustic-wave-based monitoring techniques. For plate-like structures, such guided acoustic waves can propagate over distances of several metres, offering a great deal of potential for mapping defect locations over large areas. Doing this has remained a challenge, though.
Scientists have understood the modes of acoustic wave propagation through simple structures since the 1920s, but mapping and quantifying what happens when waves are scattered by complex discontinuities, such as bends or welds, is very involved. So Dr Wilcox and colleagues have been analysing the exact physics of the interaction of the waves and structural features, which hasn't been properly tackled before. While powerful computers have made it much easier to numerically calculate and simulate complex wave behaviour, the reality is that sending a burst of guided waves into the structure results in a very messy signal.
The problem is in discriminating between very small changes due to damage and very big signals caused by the waves bouncing off structural features, says Wilcox. While digital signal processing methods can help to pre-process the signals to increase the signal to random noise ratio, the main challenge is dealing with environmental effects - particularly those due to changes in temperature.
"Changes of even a fraction of a degree can lead to residual signals from structural features that are an order of magnitude bigger than the signals you are trying to see due to a typical crack," explains Wilcox. Without performing some sort of compensation, the detection threshold has to be set so high that signals from critical defects are missed. At a lower threshold, you get an unacceptably high number of false positives.
Wilcox and colleagues have developed an intelligent subtraction algorithm that compensates for the effects of environmental changes, such as temperature and loading, in order to make the comparison of signals as stable and robust as possible. Temperature has the most dominant effect, as any increase in temperature will decrease the wave velocity enough to ruin the measurements. Loading in the structure also alters the wave velocity, depending on wave mode and direction of loading.
The technique works with a minimum of one sensor per square metre (typically a piezoelectric device that both triggers the wave and detects the scattered result) and is suitable for metal or composite structures of the order of metres and upwards in size, which might include aircraft wings, shipping containers, power station components and even wind turbines. Defects such as 5-10mm long cracks, corrosion patches, and delaminations in composites can be detected in this way.
In five years' time, Wilcox hopes that this kind of system will be ready for use on unmanned offshore production facilities where it could be used to find, say, localised corrosion.
In this research the sensors were hard-wired but there is no reason why they couldn't communicate wirelessly. "In theory you could remotely trigger the acoustic wave pulse so it could be a completely autonomous system that made measurements every hour, say, and it only sends an alarm if there's a major change," suggests Wilcox.
In the short term, permanent structural health monitoring systems may struggle for acceptance in industries that use well-established manual non-destructive testing and evaluation techniques. However, the scope for use on structures where there is little or none of this kind of testing, such as on America's bridges, in shipping containers and new types of infrastructure such as wind turbines could be significant.
There are nagging questions, though, not least the practicality of long-term installations using commercial semiconductors, which usually aren't built to last more than five years these days. That's a wider sustainability problem that will also have to be tackled, as our national infrastructure becomes increasingly virtual as well as physical.