Computer science

Street smarts are a vital ingredient in de-icing technology!

Light flurries are predicted in the next few hours along the Eastern Townships autoroute. The temperature is -11oC but the thermometers in the pavement indicate that the temperature of the asphalt is -15oC to -13oC. Yves Fournaise, foreman at the Richmond Service Centre, has to make a quick decision: should he spread salt, sand or calcium on the roads? “Today, we chose a mixture of two parts sand and one part salt,” he says.

Until now, this calculation was always done on the cuff, based on a one-page chart and, to a large extent, the experience of the team in place. But two computer experts at the Université de Montréal, Michel Gendreau and Jean-Yves Potvin, have put artificial intelligence to work for the Richmond team.  Researchers at the Centre for Research on Transportation (CRT) have developed a unique software tool that will help managers use abrasives in a way that will reduce the environmental impact, while bringing down costs. Referred to as the “neuron network,” this system is a computerized brain that takes into account a large number of variables before proposing a solution.

“De-icing and snow removal are much more complex than one might think,” explains Mr. Gendreau, Director of the CRT. De-icing experts must take into account multiple variables including traffic volume, rush hour and the road conditions. When freezing rain hits, failure to intervene adequately can end in tragedy. Experts must be prepared before the slightest meteorological disturbance occurs and must know the effects of each abrasive. “All winter long, and across Quebec, this exercise is repeated hour after hour.  It’s an art and a technique.”

And the costs are nothing to sneeze at: $11 for a ton of sand, $70 for a ton of salt, and a staggering $700 for a ton of calcium. In short, the software processes 21 factors, only some of which are meteorological in nature. The software uses a “reasoning” process based on the number of heavy trucks on the road, the objective sought (de-icing, ensuring safety or simply keeping the network operating), road conditions (is there snow, slush, hardened snow, freezing rain, black ice?), the date (are the days short or long?), etc. In terms of the weather, it’s important to know if there will be winds in the next six hours, if accumulation or a drop in temperature is predicted, etc.

The software is gradually being implemented in Richmond’s regional road network which spans 140 km and will be in full operation by next winter. The Eastern Townships is an ideal location to implement the software since it offers good geomorphologic diversity and several microclimates. But why use a computer system to manage the spreading of abrasives? “There are many advantages,” explains Mr. Fournaise. “To begin with, it will standardize maintenance interventions. At the moment, there are as many ways to calculate the proportions of salt and sand as there are foremen. Secondly, it will ensure the preservation of our expertise.”

As is the case for all of Quebec’s labour force, road maintenance experts are ageing and there is concern that their expertise will disappear when they retire. “We had to find a way to ensure that the experience acquired would be preserved,” explains Mr. Fournaise. The environment is another point of interest. The federal minister for the environment has stated that abrasives “pose a serious threat to animals, plants and aquatic environments situated close to storage and elimination sites, and due to run-off into watercourses and rivers.” The software could reduce the quantity of abrasives by 10 per cent in the first year and, therefore, help to protect the environment. A project worth following…