Early structural anomalies identification allows to hold maintenance activities that avoid loss of both economic resources and human life. This is extremely important for crucial infrastructures like railway bridges. This paper illustrates the structural health monitoring approach applied to a simply supported prestressed concrete railway bridge. In the framework of long-term monitoring, both static quantities (displacements, strains, and rotations) and environmental measurements (temperatures) have been recorded. Machine learning techniques, Extreme Gradient boosting machine and Multi-Layer Perceptron, have been exploited to build regression correlation models associated with the undamaged structural condition after adequate pre-processing operations. In this way, alarm thresholds based on the expected residuals between the predicted structural quantities and the measured ones, have been defined. The thresholds turned out to be able to catch early-stage anomalies not pointed out by traditional damage thresholds based on the design values. The proposed damage index is chosen as the moving median of the residuals, allowing a significant reduction of false alarms. The used correlation models and the obtained results represent a starting point for the generalization of this approach to the bridges belonging to the same static typology.

This paper proposes a hybrid technique to solve the inverse problem of damage localization and severity estimation in beam structures. The first phase of the method involves the use of influence lines (ILs) to extract information about the damage location. Then, a genetic algorithm (GA), representing the core of the whole procedure, uses static parameters as displacements and rotations at a few points to evaluate the bending stiffness along the structure by updating a finite-element model. The information obtained in the first phase is used in the second phase for (1) reducing the number of design variables of the GA and the consequent computational time; and (2) improving the accuracy of GA solutions because it allows a suitably trained neural network to select proper values for the coefficients of the proposed cost function in the genetic algorithm. The procedure is applied to a test problem, a simply supported, prestressed concrete railway bridge located in northern Italy. Numerical experiments are also conducted to test the procedure when the beam length and geometric properties vary.

In recent years, great attention has focused on the development of automated procedures for infrastructures control. Many efforts have aimed at greater speed and reliability compared to traditional methods of assessing structural conditions. The paper proposes a multi-level strategy, designed and implemented on the basis of periodic structural monitoring oriented to a cost- and time-efficient tunnel control plan. Such strategy leverages the high capacity of convolutional neural networks to identify and classify potential critical situations. In a supervised learning framework, Ground Penetrating Radar (GPR) profiles and the revealed structural phenomena have been used as input and output to train and test such networks. Image-based analysis and integrative investigations involving video-endoscopy, core drilling, jacking and pull-out testing have been exploited to define the structural conditions linked to GPR profiles and to create the database. The degree of detail and accuracy achieved in identifying a structural condition is high. As a result, this strategy appears of value to infrastructure managers who need to reduce the amount and invasiveness of testing, and thus also to reduce the time and costs associated with inspections made by highly specialized technicians.