Resumen de Gestión predictiva de infraestructuras acuáticas en entornos urbanos resilientes

Marc Ribalta Gené

• The sewage network is crucial for collecting and transporting wastewater to treatment plants, preventing stagnation and related health, ecological, and economic issues. However, sewer systems infrastructure degradate when transporting wastewater across time. Maintenance is essential but costly, with the need of investing heavily in recent years.

  Recent advances in machine learning allow for predictive maintenance, learning from historical data to anticipate issues like sediment accumulation. This approach targets critical areas, reducing unnecessary maintenance. However, challenges include the complexity and variability of sewer sections and citizen behavior.

  This thesis aims to enhance machine learning methods for sewage management, focusing on Barcelona. It includes four studies: identifying effective predictive objectives and algorithms, analyzing current model development issues, presenting a replicable predictive architecture, and evaluating data impact on predictions. The research demonstrates improvements in prediction accuracy and efficiency, highlighting the importance of advanced data processing and structuring techniques.