Network noise reveals need for maintenance
The reduction in travelling was initially achieved by using existing monitoring data in a new way. TDC NET constantly measures the cable network and can thus see when changes in electrical voltages occur.
"We call it network noise and have developed an AI model that can predict when network maintenance is needed based on the noise," says Mads Bossen Hansen.
Noise can occur for many reasons. It could be changing weather conditions or something as simple as the cables in the cabinets connecting the network wires are located near a road with heavy traffic, which causes vibrations. Over time, these vibrations or weather conditions can loosen the cables, eventually resulting in the connection being degraded or lost completely.
"Loose wires cause changes in the voltages. We can use this discrepancy to predict when it is necessary to send out a technician to tighten the cables before they become completely loose and the connection is degraded or interrupted and reported by one or potentially many customers," says Mads Bossen Hansen.
TDC NET came up with the idea for the model, which was further developed together with DTU. The next major task was to implement the solution in TDC NET.
"Previously, our technicians were used to driving to tasks only when a customer had reported an internet connection fault, for example. But now they must go to tasks where a fault has not yet occurred. This means that the planners and technicians have to change their mindset, and they also must note down what they repair, so we can gather experience and use it for future maintenance tasks," says Mads Bossen Hansen.
Today, TDC NET has fully embraced the new model where problems are solved before they occur. The new approach to maintenance is part of the reason why the need for driving has been reduced and why fewer customers experience network outages today.
Division into pizza slices to further reduce driving
Another result of the collaboration with DTU can be used when TDC NET soon gets a new IT system.
Based on historical data, DTU has developed a mathematical model for how it will be possible in the future to limit the part of TDC NET's driving that consists of service tasks. In other words, not the traffic for acute outages, but for customers who, for example, change network provider or want a new TV package.
"We have merged data for driving to service tasks with emergency tasks that occur during the course of a day, and which must be solved by the same technicians," says Professor David Pisinger, DTU.
"The data is used to calculate a solution that divides a region into different slices like a pizza or sliced bread. In one slice, for example, service tasks would only be performed on Mondays, in another slice on Tuesdays, etc. At the same time, the slices should be divided so that it is always possible to cover emergency tasks in the entire 'pizza area'".
Puzzle of tasks to consider
The researchers used Funen as a test case. They carried out many simulations of possible slice divisions, which were tested and subsequently optimised. New simulations, optimisations and so on followed, until they had refined the algorithm that can indicate the best way to slice an area, in this specific test, Fyn.
"By dividing a region into slices, we can reduce driving by around 10 per cent by simply changing how often the service tasks are performed - without affecting the driving for emergency tasks. This may sound like a simple solution, but it required many months of driving simulations to design it. The system must also handle the fact that service calls in a slice area suddenly must deal with emergency breakdowns and then pick up where they left off. It's a huge puzzle to solve," explains David Pisinger.
The algorithm developed is designed so that TDC NET can incorporate historical data from driving in all areas of Denmark when they get a new IT programme and thus incorporate the right slices for maintenance tasks in the individual regions.
Mads Bossen Hansen is looking forward to being able to implement this result of the project.
"It's amazing how we've been able to get this far by focusing on reducing mileage and incorporating the latest research. We still hope to fulfil our ambitious goal of cutting 25% of our total driving with the help of the solid research that has been done. From the project as a whole, we also hope that this research can help other companies and institutions that have larger fleets of vehicles to reduce their driving."
