Browsing by Author "Fernández-Navamuel, Ana"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Gradient-Boosting Applied for Proactive Maintenance System in a Railway Bridge(Springer, Cham, 2021) García-Sánchez, David; Iglesias, Francisco; Diez, Jesus; Piñero, Iñaki; Fernández-Navamuel, Ana; Sánchez, Diego Zamora; Jiménez-Fernandez, José Carlos; Rizzo, Piervincenzo; Milazzo, Alberto; Tecnalia Research & Innovation; E&I SEGURAS Y RESILIENTESThis article contributes in the research direction of the application of Machine Learning techniques in bridge safety assessment and it lays basis to further improve the accuracy of safety assessment including analysis of real data. The communication puts forward the process and model of scale measured points correlation of bridge monitoring system on the frequency domain as a tactic to control the influence of a railway device (crossing) located on the top deck of a railway bridge. The process and model are put forward mainly for the characteristics of the damage detection for long-term assessment, going from an intensive multi-sensor monitoring system to a softer one. Finally, a Gradient-Boosting multi-regressor method has been developed to be easily implemented in a warning system that provides predictive skills to the current preventive maintenance strategy. The method is validated by simulating the undamaged and abnormal scenarios with Monte Carlo method.Item Vibration-Based SHM Strategy for a Real Time Alert System with Damage Location and Quantification(Springer Science and Business Media Deutschland GmbH, 2021-01-11) Fernández-Navamuel, Ana; Zamora-Sánchez, Diego; Varona-Poncela, Tomás; Jiménez-Fernández, Carlos; Díez-Hernández, Jesús; García-Sánchez, David; Pardo, David; Rizzo, Piervincenzo; Milazzo, Alberto; E&I SEGURAS Y RESILIENTES; Tecnalia Research & InnovationWe present a simple and fully automatable vibration-based Structural Health Monitoring (SHM) alert system. The proposed method consists in applying an Automated Frequency Domain Decomposition (AFDD) algorithm to obtain the eigenfrequencies and mode shapes in real time from acceleration measurements, allowing to provide a diagnosis based on a Support Vector Machine algorithm trained with a database of the modal properties in undamaged and damaged scenarios accounting for temperature variability. The result is an alert system for controlling the correct performance of the structure in real time with a simple but efficient approach. Once the alert is triggered, the undamaged mode shapes (which could be previously stored in a database of modal parameters classified by temperature) and the current (damaged) mode shapes, can provide guidance for further application of Finite Element Model Updating (FEMU) techniques. The method is trained and validated with simulations from a FE model that is calibrated employing a genetic algorithm with real data from a short-term vibration measurement campaign on a truss railway bridge in Alicante (Spain).