Structural System Identification of Nonlinear Energy Sink with Negative Stiffness Using Fourier Neural Network

The identification of structural systems using vibration measurements has garnered a lot of interest in the field of structural dynamics over the last few decades. System identification is the process of modelling dynamical systems mathematically using measurements of input and output signals. Extended Kalman filter (EKF) or unscented Kalman filter (UKF) are commonly used to identify the structural properties. EKF cannot estimate the state and parameters of the structure when nonlinearity is present in the system. On the other hand, UKF takes a lot of computational time for state and parameter estimation. With this in view, in this study, a Fourier neural network (FNN) has been used to solve the identification problem. A neural network establishes a mapping between infinite-dimensional spaces. For state estimation, FNN uses the advantages of Fourier transformation. In this study, a single-story steel moment-resisting frame coupled with a nonlinear energy sink with nonlinear stiffness is considered for the numerical demonstration purpose. Results show the effectiveness of the proposed approach, whose performance is compared with UKF in terms of accuracy, robustness to noise level, and computational efficiency.