An innovative framework for structural health monitoring of long-span bridges

Structural Health Monitoring of structures has received a wide diffusion and development in the last decade. Steps forward have been made in many fields as, for instance, in sensing technologies, data migration facilities and data pre/post processing algorithms. Many of these advancements have found their first application on monitoring of bridges which are considered among the most crucial structures to be monitored because of both their importance in the economic activities of a country and the relevance of the economic effort necessary for their construction and maintenance. Within this context the structural health monitoring of bridges aimed at the continuous condition assessment of such structures not only seems to be very promising for increasing the cost-effectiveness of the maintenance procedures but can be still considered a challenge. To this purpose tools based on multivariate statistical analysis are becoming very popular for automatically revealing the existence of damage in structures using vibration data under changing environmental and operational conditions (typically temperature, humidity and traffic intensity). In the present PhD Thesis, considering natural frequencies as damage-sensitive features, multivariate statistical analysis are newly applied for monitoring the structural health state of bridges, accounting for the linear and nonlinear correlations between such dynamic features and the environmental and operational conditions. A procedure based on the continuous modal frequencies tracking, Principal Component Analysis and Novelty Detection is proposed. The effectiveness and the capability in damage detection of such technique is previously tested on the pseudo-experimental response data of an analytical parametric model of suspension bridge with damage in one main cable and subjected to wind loading and changing temperature. Thereafter, in order to have an absolutely realistic representation of the operational and environmental conditions, the same technique is tested on long-term real bridge data. The obtained results demonstrate in both cases the feasibility of permanent monitoring systems for the real-time condition assessment of bridges and the robustness of the proposed procedure in revealing the existence of damage. In fact, even if the effects induced on frequencies by damage are very small, of theorder of few per mil, in any case smaller than those induced by environmental and operational conditions, the adopted statistical technique allows to reveal its occurrence in a reliable and prompt manner. Moreover, the absolute general nature of the proposed approach may reveal, in perspective, its extension to any other structure equipped with a permanent monitoring system.