Automatic operational modal analysis: challenges and applications to historic structures and infrastructures

The core of the work turns around the capability to automate Operational Modal Analysis methods for permanent dynamic monitoring systems. In general, the application of OMA methods requires an experienced engineer in experimental dynamics and modal analysis; in addition, a lot of time is usually spent in manual analysis, necessary to ensure the best estimation of modal parameters. Those features are in contrast with permanent dynamic monitoring, which requires algorithms in order to efficiently manage the huge amount of recorded data in short time, ensuring an acceptable quality of results. Therefore, the use of parametric identification methods, like SSI methods, are explored and some recommendations concerning its application are provided. The identification process is combined with the automatic interpretation of stabilization diagrams based on a damping ratio check and on modal complexity inspection. Finally, a clustering method for the identified modes and a modal tracking strategy is suggested and discussed. The whole procedure is validated with a one-month and a one-year set of "manually-identified" modal parameters. This constitutes a quite unique set of validation data in the literature. Two monitoring case studies are studied: a railway iron arch bridge (1889) and a masonry bell-tower (XII century). Within this framework, classical and new strategies to handle the huge amount of recorded and identified data are proposed and compared for structural anomaly detection. The classical strategies are mainly based on the inspection of any irreversible frequency variation. To such purpose, it is mandatory an extensive correlation study with environmental and operational factors which affect the frequency of the vibration modes. Conversely, one of the proposed strategy aims to use alternative dynamic features that are not sensitive to environmental factors, like mode shape or modal complexity, instead of frequency parameters in order to detect any structural anomaly. In addition, a further strategy has the goal to eliminate the environmental-induced effects on frequency without the knowledge and the measurements of such factors. The procedure is mainly based on the combination of a simple regression model with the results obtained by a Principal Component Analysis. Furthermore, two automated Operational Modal Analysis (OMA) procedures are compared for Structural Health Monitoring (SHM) purposes: the first one is based on SSI methods, while the second one involves a non-parametric technique like the Frequency Domain Decomposition method (FDD). In conclusion, a model updating strategy for historic structures using Ambient Vibration Test and long term monitoring results is presented. The main goal is to integrate the information provided by a FE model with those continuously extracted by a dynamic monitoring system, basing so any detection of structural anomalies on the variation of the uncertain structural parameters.