An application of data-driven analysis in road tunnels monitoring

In order to comply with the minimum safety requirements imposed by the Directive 2004/54/EC it is of paramount mportance to correctly manage the operation and maintenance of road tunnels. This research describes how Artificial Intelligence techniques can play a supportive role both for maintenance operators in monitoring tunnels and for safety managers in operation. It is possible to extract relevant information from large volumes of data from sensor equipment in an efficient, fast, dynamic and adaptive way and make it immediately usable by those who manage machinery and servicesto aid quick decisions. Carrying out an analysis based on sensors in motorway tunnels, represents an important technological innovation, which would simplify tunnels management activities and therefore the detection of any possible deterioration, while keeping the risk within tolerance limits. The idea involves the creation of an algorithm for the detection of faults by acquiring data in real time from the sensors of tunnel sub-systems and using them to help identify the service state of the tunnel. The AI models are trained on a period of 6 months with one hour time series granularity measured on a road tunnel part of the Italian motorway systems. The verification has been done with reference to a number of recorded sensor faults.