A breakthrough in predicting where potholes will appear on our roads should help to cut the massive cost of repairing them, writes Dick Ahlstrom.
THE IRISH POTHOLE is a mysterious thing, appearing as out of nowhere to the annoyance of road users and council officials alike. An engineer has intervened to take the mystery out of the pothole, however, by devising a way to predict where they are likely to occur.
A serious amount of number crunching was involved, but University College Dublin's Prof Eugene O'Brien now has a way to identify at-risk road surfaces. His methods should allow road crews to carry out preventative repairs at a much lower cost before full road resurfacing becomes necessary.
"It is a big breakthrough but it is complicated," states Prof O'Brien, professor of civil engineering in UCD's School of Architecture, Landscape and Civil Engineering. It relies on Bayesian statistics and mathematical models which can predict where damage will appear on a road surface, and give an indication of when a road surface is likely to fail.
He came to the subject in a roundabout way, starting with his involvement in bridge engineering, and later with attempts to measure the weight of trucks while they are moving. "I am a bridge engineer and very interested in [ weight] loading on bridges, but there was little funding for research, so I started looking at truck weigh-in-motion."
Bridges lend themselves to this as sensors can be placed under bridge structures or on road surfaces. The work involved is very important, given that overweight trucks are a serious road hazard and damage road surfaces.
"Heavily overloaded trucks are a risk because they become unstable and are quite dangerous," he says. They also waste taxpayers' money. "If you take one overloaded truck driving on a hot day, it can cause millions of euro in damage."
Being able to assess weight while a lorry is on the move would help efforts by gardaí to prevent overloading. "The holy grail of weigh-in-motion is automated enforcement," he says.
Such a system would detect an overloaded lorry and alert police further down the road. "You need to take a static weight to prove it," Prof O'Brien explains. Both France and the Netherlands are developing weigh-in-motion enforcement systems.
Many such systems rely on surface sensors based on piezoelectricity, where pressure on a crystal produces an electrical signal. Prof O'Brien has been using data from a system that provides information from 16 sensor points. "With a lot of complex [ mathematical] algorithms you can work back to the weight on each of the axles," he says.
Working with student Abraham Belay and University of Nottingham colleague Dr Andrew Collop, he found the sensors provided a much wider range of information related to how a truck was "bouncing and rocking" on the road surface.
They analysed the data and realised that the vehicles were exerting forces on the road surface in an unexpected way. "The pattern of force that the truck applies to the road repeats and can be predicted," says Prof O'Brien.
They also found that two trucks, when compared, could have very different force patterns, but averaging results from many trucks allowed them to develop a statistical model of the repeating pattern.
This was an important finding. "The assumption is that each part of the road is as susceptible to damage as the next, but this is not so," he explains. Damaging stress might occur at one point but skip adjacent areas. The force exerted depends on two things, the "road surface profile" and the characteristics of a given truck.
The latter relates to things such as the stiffness of springs, suspension dampening, tyre stiffness and other attributes. He blended these characteristics to provide a way to understand where peak force was being applied on the basis of the road-surface profile. "We have now found a way to predict those patterns using a Bayesian statistical method."
This, in turn, helped them to build a "road damage model" that can indicate not alone where subsurface fracturing will occur, causing breakdown, but also give an indication of how long it will take for the damage to appear.
"It takes different lengths of time to damage the pavement," he states. "What is most exciting is we should be able to detect the patterns early."
Once damage begins to occur, it changes the road surface profile. A feedback begins to occur, accentuating later damage. "Once the pattern becomes self perpetuating, we should be able to predict that," Prof O'Brien says.
Annoyingly, virtually all the surface breakdown on our roads is due to overweight lorries. "Cars make no difference at all, it is all about trucks," he says.