STUD:Techrail - Modern Tunnel Inspection With Machine Learning

The technology used for Mars exploration can keep our tunnels safe. Norway is one of the countries in the world with the most tunnels. In fact, there are only two countries in the world with more tunnels than us: Japan and China. One might think that a tunnel is something that is built once, and then remains open for ever and ever, without furher intervention. This is not the case. The mountains are constantly moving, cars drive through the tunnel and wear it down, and there might be water running in the tunnel. It is therefore important to routinely check the state of the tunnel to verify that it is fit for use. In Norway road tunnels are checked every five years by tunnel inspectors. Checking the tunnel can be a cumbersome process. A tunnel expert must walk through the tunnel by foot and look for maintenance needs. The cases that require maintenance can be a rusty bolt or air pockets forming between the concrete wall in the tunnel and the mountain behind it. During the tunnel inspection the tunnel has to be closed. We want to facilitate this process. We have mounted a Lidar - An advanced scanning device with high precition - on top of a car. With the Lidar mounted on our car we drive through tunnels at regular speed and scan the tunnel. From the collected data we build a model for what the tunnel looks like and apply machine learning algorithms to discover maintenance needs. When we drive through the tunnel again at our next inspection we can compare the model with today's result, and see where the tunnel has changed. In order to demonstrate the use cases of our method, we have gathered data from Gimstunellen in Trøndelag. We visualized the tunnel and made a python function that finds the width of the tunnel at a given height. There are many benefits of this new method. We do not need to close the tunnel for inspection, and we can use historical data to get a better grasp of what is going on in the tunnel. We hope to leverage this to discover maintenance need even earlier than what the tunnel experts can do today, and make sure that your journey through the tunnel is as safe as possible. Through our collaboration with the research council we were able to prove that gathering data from a tunnel using a sensor platform mounted on a car is a viable option for detecting topological changes in the tunnel. Market actors have shown interest in our project and a willingness to pay for the service given that the end product works as intended. We hope that in the future other companies will be inspired by our work and provide an end-to-end solution that facilitates tunnel inspections to a level where there is no need to shut down the tunnel. This will allow for more tunnel uptime and increased safety, as maintenance needs are detected earlier and more accurately.

Oppnådd effekt: Prosjektet viser at det er mulig å samle inn data av en tunnel fra en kjørende sensor-plattform og bruke den i analyser til å oppdage topologiske endringer i tunnelen. Vi hadde også god dialog med bransjeaktører som viste stor interesse/kjøpsvilje gitt at produktet virker. Potensiell effekt: I fremtiden kan det være at vi ser andre aktører fra bransjen ta inspirasjon fra dette prosjektet og ferdigutvikle det til et komplett produkt hvor tunnelen slipper å stenges ned. Dette vil føre til tryggere tunneler med mer oppetid da en kan fokusere vedlikeholdet på områder av interesse som kommer frem av skanningen.