Civil infrastructures, such as bridges and viaducts, have always been prone to various types of damages due to environmental and operational factors.
There is therefore a pressing need of suitable maintenance strategy in order to permanently ensure the safety and reliability of structures.
Currently the most common way to perform civil infrastructure maintenance is periodical visual inspection. One main drawback of this conventional approach is the damage that can develop and left undetected in between the inspection periods, that can lead to potential catastrophic consequences. Moreover, this method requires highly-trained technicians
and easy access to the monitored structure parts, which is seldom the case for bridges currently open for traffic.
In this frame, structural health monitoring (SHM) has emerged as a promising damage detection strategy. SHM implements continuous monitoring systems to detect failures at the earliest possible time, to obtain real-time information after disasters and extreme events such as earthquakes, and provide hints for planning inspection, maintenance and repair of the structure. Yet there are still many challenges to implement SHM as the monitoring of large infrastructures requires a plurality of transducers (accelerometers, strain gauge, ...) arranged in complex and costly architectures.
In this frame, fiber optic sensors have recently proved to be an interesting alternative for SHM applications. Several fiber optic systems defined as "distributed sensors" can actually provide a detailed monitoring of the whole structure with spatial resolution of meters along sensing length up to many kilometers. On the other hand, a simpler and more cost-effective approach that can however provide an equally useful information, is represented by "integral measurements". This approach proves to be particularly suitable to perform a dynamic analysis of the modes of a structure and to analyse how these modes change in presence of structural failures.
In the present PhD work both these approaches will be taken into account to perform SHM pilot activities on different civil infrastructures (viaducts, bridges, river banks, ...). At first, the work will foresee an initial phase of theoretical and technical in-depth analysis and comprehension of the above mentioned fiber optic sensing solutions. This phase will be propaedeutic in order to carry out further technological advances to achieve a suitable trade-off between performance and costs that could make these solutions reliable and affordable for many SHM applications.
Besides, a further aspect that will be addressed during the PhD is related to the analysis and identification of methodologies related to the deployment of the sensing fiber cable. This aspect represents a key issue if fiber-optic sensors are to become a well-established and widely used technology for SHM applications. Indeed, the sensitivity of the whole monitoring system is closely linked to the capability to correctly transfer the deformations of the structure to the sensing fiber cable. Proper fiber coatings and adhesion/anchoring systems will be developed to ensure a good transfer and, at the same time, a reliable protection from chemicals, rodents and any extreme environmental conditions.
The performance of the identified solutions will be first experimentally evaluated on test-rigs (bridge cranes, landslide emulators, ...) available in the department laboratories of the Politecnico, carrying out both static and dynamic deformation measurements. Subsequently, a final in-field validation phase will be foreseen where structural health monitoring will be carried out on real civil infrastructures.
A further challenge of this PhD will be related to the interpretation of all the data acquired by the installed sensors and how to convert them into useful information for carrying out SHM and maintenance strategy. Automation of this data processing steps will be also an important issue to be addressed especially when dealing with a huge set of data as those provided by fiber optic sensors. In this context, machine learning emerges as a solution to automate the pattern recognition problem, as it exploits mathematical-computational methods to learn information directly from the data even without mathematical models related to the monitored infrastructure.
Thus, a further activity that will be carried out during this PhD will be to combine the capabilities of both “distributed” and “integral” fiber optic sensors with the potentiality provided by machine learning in order to achieve an innovative and reliable diagnostic systems for structural monitoring applications.