Bridges and viaducts are key elements of any land transport infrastructure, allowing for shorter and more economical journeys by avoiding the need to bypass obstacles such as rivers and valleys. This aspect is even more important when it comes to the railway in which vehicles are not able to overcome very intense slopes. Recent technological advances in trains have led to significant increases in speeds and axle load significantly increasing demands on bridges and viaducts. These increases in demand, linked to the aging of the infrastructure, are making it increasingly important to have robust methodologies for the early detection of damage in these structures. Nowadays, the standard, both in Brazil and in the world, is still the performance of purely visual periodic inspections, which does not allow the early detection of many types of damage. Faced with this reality, some infrastructure managers have chosen to install modern structural integrity monitoring systems on bridges. The high cost associated with these systems, however, ends up making their installation on all bridges and viaducts of a railroad prohibitive, limiting their use to those that are more critical or of greater length. In an attempt to address this deficiency, infrastructure managers are increasingly interested in new technologies capable of detecting road damage based on monitoring systems embedded in rail vehicles. Faced with this motivation, this work proposes the development of an automatic methodology for the detection of damage in railway bridges and viaducts based on the recognition, through machine learning techniques, of characteristic patterns of damage in dynamic responses obtained through in-vehicle monitoring systems.
News published in Agência FAPESP Newsletter about the scholarship: