Parametric modelling of dynamic structure may benefits from features of particle swarm optimization (PSO), which robust and fast in solving nonlinear, non-differentiable, and multimodal problems. This paper presents the PSO approach which includes an improved algorithm to model a flexible plate structure parametrically. The introduction of a dynamic spread and momentum factor, both by modifying the inertial weight of each particle, solves the problem of getting stuck at local optima, preserves diversity and trades-off between exploration and exploitation. The identification is performed on basis of minimizing the mean-squared output error and is assessed with correlation tests and in time and frequency domains. It is shown input-output characterization in time and frequency domains that the improved algorithm possesses features of accuracy and quick convergence.