BRIDGES MONITORING: AN APPLICATION OF AI WITH GAUSSIAN PROCESSES

Abstract

. Monitoring and assessment of existing infrastructures is gaining attention in nowadays society as aging degradation reduces the residual structural safety. This work aims at providing a competitive procedure for the monitoring of bearings of bridges, based on data analysis coming from the use of strain gauges. It proposes an unsupervised learning (USL) approach based on a single-valued performance indicator to monitor the global behavior of bearings as critical components in a viaduct. The procedure for data analysis is exploiting the well-known Gaussian Processes to analyze and correlate horizontal bearing displacements and environmental temperature, as primary source of deformation of the deck. With this approach, we combine measurements from twelve sensors to design a robust performance indicator that is only weakly affected by temperature variations. Data exploitation and manipulation is described, from data acquisition and selection to the interpretation of results. Meanwhile the availability of a huge amount of information coming from the monitoring of big structures is no more an issue, the management of such indications is not always clear and efficient. For this reason, the method proposed by this paper can be of support in dealing with real-time monitoring of bridge bearings, maximizing unexpected maintenance and increasing the safety of the bridge.