Investigations into aerodynamics of tall buildings using signal processing tools

¹ Civil Engineering Department, National Institute of Technology Hamirpur, Hamirpur - 177005, Himachal Pradesh, India
^2,3,4 Wind Engineering Lab, CSIR-Structural Engineering Research Centre, CSIR Campus, Taramani, Chennai - 600113, Tamil Nadu, India

Abstract

In rapidly expanding urban environments, the construction of tall, sky-piercing buildings is essential. These slender structures have low resonant frequencies and minimal damping, making them particularly susceptible to lateral loads, such as wind load. Understanding wind-induced loads and building responses remains in a relatively nascent stage within structural engineering. This paper investigates surface pressures obtained from wind tunnel testing of a CAARC model building using advanced signal processing techniques, including Principal Component Analysis (PCA) and Independent Component Analysis (ICA). By applying PCA, six dominant pressure modes of the were identified, capturing 80% of the total variance representing the dominant dynamic behaviour while reducing dimensionality. This reduction process is essential for simplifying complex and random field data sets while preserving significant patterns and trends. ICA further isolated independent sources of variability, providing insights into the underlying physical processes affecting the building. The ability of ICA to separate these independent components allows for a clearer understanding of the individual influences of different wind load factors. Through comprehensive PSD analysis, we identified dominant frequencies associated with the wind load factors, linking them to specific wind load mechanisms and their effects on building oscillations. PSD results were consistent with existing literature, confirming the presence of low-frequency oscillations around 10 Hz, which are characteristic of vortex-induced vibrations in tall buildings. This comparison highlights the accuracy and applicability of our methods, reinforcing the potential of these techniques for improving predictive models.