A novel recursive dynamic mode decomposition method for modal parameter identification of time-varying structures based on input-output data

Many engineering structures under operating conditions exhibit significant time-varying dynamic characteristics, making time-varying modal parameter identification an urgent but challenging problem in structural dynamics. In recent years, the Dynamic Mode Decomposition (DMD) method, as a promising data-driven analysis approach, has received increasing attention and favor of scholars and engineers in the field of structural dynamics. This paper aims to propose a novel recursive DMD method for modal parameter identification of time-varying structures based on input-output data. On the one hand, the traditional output-only DMD framework is extended by considering system inputs to improve the modal parameter identification accuracy. On the other hand, the recursive format of the DMD framework is proposed by integrating the Projection Approximation Subspace Tracking (PAST) algorithm, allowing the proposed method to operate in a recursive manner with better computational efficiency. The proposed method is, respectively, validated through a numerical system with time-varying stiffness and an experimental system with abrupt-changing mass. The identification results demonstrate that the proposed method can accurately identify and track the time-varying modal parameters of structural systems based on input-output data. Moreover, the performance of the proposed method can be further improved by tuning the DMD model parameters and selecting appropriate types of vibration response data, which is of great significance for identifying time-varying modal parameters in practical applications.