Advanced monitoring of railway infrastructure
|Project members:||dr.ir. K.A.J. Verbert (Kim), prof.dr.ir. B. De Schutter (Bart), prof.dr. R. Babuška (Robert)|
|Keywords:||Identification and estimation, Machine learning, Railway networks|
|Sponsored by:||STW, Prorail|
Prevention of disruptions and minimization of costs are of paramount importance for railway infrastructure managers like ProRail in order to guarantee robust performance of the railway network and to satisfy requirements from customers and government. Therefore, ProRail needs an information and monitoring system that can detect, localize, and diagnose disruptions and emergent problems in a fast and efficient way, and that can propose preventive maintenance, repair, or replacement actions. In this way, effects of incidents and disturbances will eventually be minimized and predictive repair or maintenance actions can be optimized, resulting in minimum disruptions and costs. Various measurement and detection methods have been employed and new technologies are being developed to monitor the condition of the railway infrastructure. Currently, a large amount of measurement and management data are available at different sources. To make use of all relevant available data an integrated decision support system is therefore desired. Such a system should be intelligent and it should make use of the increasing number of sensing, measuring, communication, and control equipment present in today's railway infrastructure.
This project will develop new, intelligent, systematic, efficient, and robust methods for monitoring of railway infrastructure, i.e., methods that are continuously fed with new data collected from the various measurement units in the railway network and that continuously monitor these data and raise an alarm as soon as an actual or emergent problem is detected. The main challenges to be addressed are increasing the reliability, efficiency, robustness, and scope (both with respect to amounts and types of data used, missing data, and types of properties being monitored) of the monitoring process. To this aim we will combine state-of-the-art information fusion methods with new (possibly probabilistic) models for the dynamics and evolution of tracks and trains, for degradation, and for faults, as well as advanced fault diagnosis and detection methods, statistical analysis, and risk management methods. The innovative combination and integration of Artificial Intelligence (AI) methods (such as neural networks, fuzzy logic, and learning) with model-based methods will provide additional levels of efficiency and scope that cannot be obtained by a purely AI or a purely model-based approach. The project will yield efficient and effective methods for the quick detection, diagnosis, and localization of
disruptions as well as for the determination of preventive and corrective (maintenance or repair) actions to deal with these problems.