Càtedra Taigua of Urban Water Engineering and Management

Failure prediction

Predicting the probability of failures in the network should enable preventive maintenance. The use of Machine Learning for this application has been the focus of the work developed by the chair. The sector has a large amount of data, and extracting knowledge from this information is a priority.
  • Failure prediction and renewal proposals based on risk. Within the FLUENT project, in which both cs2ac and TAIGUA participate, a series of models (Logistic Regression, Random Forest, and Neural Networks) for failure prediction have been designed and calibrated using data from several companies. The failure probabilities obtained from these models, combined with the economic, environmental, and social consequences, are used to make recommendations for renewal strategies. The chair’s support through the David Alcaraz (INIREC) scholarship has allowed further exploration of the theoretical aspects of model validation and interpretation. As a result, the double bachelor's thesis (Automatic Electronic and Mechanical Engineering) Study and prediction of failures in pipes of the drinking water distribution network was completed.