ABSTRACT
The study and development of new techniques for structural monitoring have been driven by the growing number of slender structures, more susceptible to excessive vibrations, as well as by the concern about the performance and degradation of ancient structures.
With this in mind, tools with the ability to capture and interpret quickly and reliably the response of large structures become essential to complement structural damage detection techniques, especially those that are based on dynamic properties as the natural frequencies and their respective vibration modes, which are affected by the presence of defects in the structure (Adams, 1978).
In this context, in order to evaluate the performance of Self Organizing Maps (SOM) in a Structural Health Monitoring system (SHM) to detect structural damage, we used in this study the extensive monitoring database of the Infante D. Henrique Bridge (Fig. 1), in Portugal, developed by the Laboratory of Vibrations and Monitoring (ViBest) of the University of Porto (Magalhães, 2010). Infante D. Henrique Bridge (<xref ref-type="bibr" rid="ref82_3">Magalhães 2010</xref>). https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig82_1.jpg"/>
To detect changes in the frequencies of the bridge that may indicate the presence of damage, two types of analyses were performed, the first relating to the use of the estimated bridge frequencies, within both the training phase and the test phase.
Since there is no evidence of the structural anomalies in the bridge and to evaluate the functioning of the SOM, the second type of analysis was performed using the estimated bridge frequencies for training, and corrected frequencies (affecting the modal estimates by relative variations given by the numerical simulation of the damage scenarios) for testing.
The results showed a good performance of the SOM, as the implemented method was able to detect relatively small structural changes.
In this context, the achieved results highlight the utlity of nerural network in the context of vibration based structural health monitoring, since early detection of anomalies is possible.
