Principal Component Analysis of Monitoring Data of a High-Rise Building: The Case Study of Palazzo Lombardia

This chapter presents the use of principal component analysis (PCA) to extract and summarize the information contained in data acquired by a structural health monitoring (SHM) system for a high-rise building: the Palazzo Lombardia skyscraper in Milano. The structure is one of the tallest skyscrapers in Italy, and it is the house of the regional government. Due to its strategic relevance, Palazzo Lombardia is equipped with an SHM system composed of clinometers and accelerometers, spread over 5 of the 42 floors of the building, to collect information on the static and dynamic behavior of the structure over time. In previous works, empirical models were proposed to predict the eigenfrequencies of the first vibration modes of the skyscraper as a function of meaningful predictors, accounting for environmental and operational factors responsible for the structural properties changes, i.e., temperature, sun exposure, and wind direction and speed. From an SHM perspective, the predicted eigenfrequency values can be compared with the ones estimated through the adoption of operational modal analysis. Since the model is tuned on data coming from the healthy structure, a significant discrepancy between predicted and estimated eigenfrequency can point out the presence of damage. Due to the complexity and size of large-scale buildings, quantifying the influencing variables is a hard task (e.g., the temperature that dramatically changes at different locations of the structure) and, so far, other highly correlated physical quantities have been used in place of direct measurements (e.g., root mean square of the accelerations to account for the effect of wind). However, only a limited number of sensors were used, at the risk of not including other meaningful variables or additional information provided by other positions along the structure. In this work, a preliminary investigation on the use of PCA is presented, aimed at providing a few synthetic indexes, which are highly correlated with the environmental factors that mostly affect the dynamic behavior of the structure. Through the adoption of the PCA, the dimensionality of the problem can be reduced by considering only a few variables, while retaining as much as possible of the variation present in the entire data set. The potential of the proposed approach is discussed on long-term data coming from the operating monitoring system, presenting a case study in the field of SHM of real large-scale civil structures.