Contents lists available at ScienceDirect Engineering Structures journal homepage: www.elsevier.com/locate/engstruct Identification of nonlinear aerodynamic damping from stochastic crosswind response of tall buildings using unscented Kalman filter technique Yanchi Wu, Xinzhong Chen ⁎ National Wind Institute, Department of Civil, Environmental and Construction Engineering, Texas Tech University, Lubbock, TX 79409, USA ARTICLE INFO Keywords: Nonlinear aerodynamic damping Unscented Kalman filter Crosswind response Vortex-induced vibration Tall buildings ABSTRACT This study presents a new approach for identification of aerodynamic damping of tall buildings from stochastic crosswind response time history using unscented Kalman filter (UKF) technique. The system damping ratio is expressed as a polynomial function of building displacement or velocity, which is equivalent to a polynomial function of amplitude of harmonic motion. The structural system with nonlinear system damping under sto- chastic excitation is modeled as a single degree of freedom system with a white noise input. The augmented state variables of the system, which also include unknown system frequency, damping and stochastic excitation parameters, are estimated simultaneously with the UKF technique from the response measurement data. The aerodynamic damping is then extracted from the system damping by subtracting the structural damping. Firstly, the stochastic response time histories of a tall building model at various wind speeds with a known nonlinear aerodynamic model are simulated and the performance of the UKF technique and the influence of selections of various parameters involved are investigated. Secondly, the aerodynamic damping of a square-shaped tall building is identified based on aeroelastic building model wind tunnel test data. The response statistics are then computed using the identified damping and compared with the measurement data to verify the accuracy of identification. 1. Introduction The aerodynamic damping of flexible tall buildings and other structures has a negative value and exhibits apparent dependence on vibration amplitude at the vicinity of vortex lock-in wind speed (e.g., [22,21,10–13]). Accurate modeling of aerodynamic damping is criti- cally important for assessing crosswind response of these structures. The forced-vibration building model test in wind tunnel has been used to determine motion-induced, or self-excited force, whose com- ponent in phase of vibration velocity provides information on aero- dynamic damping (e.g., [31]). On the other hand, the free vibration data from wind tunnel test are often utilized to extract the amplitude- dependent aerodynamic damping associated with vortex-induced vi- bration of bridge deck sections (e.g., [14,28,26]). This free vibration method is not commonly applied for tall buildings. Research efforts have also been devoted to extracting damping from stochastic response time history data. One of this type of approaches is the random de- crement technique (RDT), which is effective for extracting linear damping (e.g., [30,20]), but is not applicable to systems with ampli- tude-dependent nonlinear damping [16]. Hao et al. [16] recently pro- posed a new approach using response standard deviation (STD) and kurtosis at different structural damping levels. On the other hand, several linear and nonlinear system identifica- tion methods in civil engineering applications have been developed, such as least-squares method, recursive least-squares method and Kalman filter (KF) approach (e.g., [17,6,37]). The Kalman filter has been widely utilized to estimate state variables and parameters of a dynamic system (e.g., [15]). The traditional linear Kalman filter has been extended to nonlinear systems with linearization of all nonlinear models, known as extended Kalman filter (EKF) technique (e.g., [32]). The unscented Kalman filter (UKF) technique has been developed for better estimations of the means and variances of state variables by using a minimal set of carefully chosen sample points (e.g., [32,19]). The applications of UKF technique to structural systems have been reported in literature (e.g., [23,34,25,7,8,35,1,4,9,24]). For example, Azam et al. [4] presented a UKF scheme for simultaneous estimations of state variables and identification of structural parameters with uncertain properties under wind loading. Lei et al. [24] proposed a UKF for si- multaneous identifications of nonlinear structure parameters and un- known excitations. This study presents a new approach with UKF technique for iden- tification of nonlinear aerodynamic damping using stochastic response https://doi.org/10.1016/j.engstruct.2020.110791 Received 26 December 2019; Received in revised form 28 April 2020; Accepted 10 May 2020 ⁎ Corresponding author. E-mail addresses: yanchi.wu@ttu.edu (Y. Wu), xinzhong.chen@ttu.edu (X. Chen). Engineering Structures 220 (2020) 110791 0141-0296/ © 2020 Elsevier Ltd. All rights reserved. T