Data‐Driven Models for Prediction of the Modal Frequencies Drift and Fluctuations as Function of the Environmental Conditions for a 50‐Story Skyscraper in a Seismic Region

ABSTRACT The effectiveness of structural health monitoring (SHM) by detecting changes in the structure's vibrational frequencies is limited by their sensitivity to factors other than damage, referred to collectively as environmental and operating conditions (EOC). The EOC produce changes of the order of those caused by small or localized damage and can mask or be misinterpreted as damage. These effects differ from one structure to another, and depend on factors, such as, for example, the construction material, age, and local climate characteristics. We present linear regression models that predict the variations with weather of the daily averages of four modal frequencies of a 50‐story reinforced concrete skyscraper, located in Yunnan province, China, a seismic area with mild climate and distinct wet and dry seasons. We use low‐amplitude accelerations and weather data recorded continuously over a period of 39 months by the building's permanent SHM system. The models are in terms of temperature, humidity, and wind speed. We also found gradual permanent reduction of the frequencies with time (1%–2% over 39 months), which we modeled as an exponential decay. We present models with constant and season‐dependent coefficients. The removal of the baseline drift and fluctuations with weather using the models lowers significantly the variability in the observed modal frequencies and the threshold of damage that can be detected from their changes.