Machine Learning-Based Structural Health Monitoring of Dams

Dams are an important asset of the European Countries in the Alpine region. Italy, for instance, hosts more than 500 large dams, initially conceived to support the development of the Country in the period between the two world wars, and in the years 1950-1970. Most of them are still in operation to produce electricity, supply water for drinking and irrigation, contribute to the mitigation of the often dramatic consequences of current climate change. The safety assessment of these strategic infrastructures, which are getting old, is supported by sensor networks that collect environmental data and response measures. The number of monitoring devices and the acquisition frequency have generally increased over time. The large amount of gathered information is usually processed through interpretation functions, while machine learning tools have recently been introduced as early recognition methods of possible anomalies in the structural response. This contribution summarizes the most recent results obtained in this context, illustrates the performance of the most promising approaches, even if not yet fully validated, discusses the still open issues and presents the latest trends.