Data-driven methods in application to flood defence systems monitoring and analysis

Modern society fully depends on properly functioning energy, communication and other types of systems. Natural or technological hazards can cause great losses if these systems are functioning improperly. Condition monitoring of objects protecting us from possible hazards becomes nowadays more and more important problem.
The problem of the flood defence structures condition monitoring is considered in this thesis. Analysis of levee behaviour using the measurements collected from the sensors installed inside the dams provides the early warning indicators. Available time before the possible levee collapse can be used for mitigation of possible effects of the flood. The goal of this work was to make a new step in investigation of the concepts of levee condition monitoring for providing on-line alerting. We developed a data-driven approach for levee condition monitoring.
We demonstrated that it is possible to monitor the levee behaviour using the data-driven models independently and in combination with physical modelling. The developed approach can be used by the domain experts as an alerting tool in order to reduce time for decision making.