HHT-based nonlinear signal processing method for parametric and non-parametric identification of dynamical systems

This paper presents a time–frequency signal processing method based on Hilbert–Huang transform (HHT) and a sliding-window fitting (SWF) technique for parametric and non-parametric identification of nonlinear dynamical systems. The SWF method is developed to reveal the limitations of conventional signal processing methods and to perform further decomposition of signals. Similar to the short-time Fourier transform and wavelet transform, the SWF uses windowed regular harmonics and function orthogonality to extract time-localized regular and/or distorted harmonics. On the other hand, HHT uses the apparent time scales revealed by the signalʹs local maxima and minima to sequentially sift components of different time scales, starting from high- to low-frequency ones. Because HHT does not use pre-determined basis functions and function orthogonality for component extraction, it provides more accurate time-varying amplitudes and frequencies of extracted components for accurate estimation of system characteristics and nonlinearities. Methods are developed to reduce the end effect caused by Gibbs’ phenomenon and other mathematical and numerical problems of HHT analysis. For parametric identification of a nonlinear one-degree-of-freedom system, the method processes one free damped transient response and one steady-state response and uses amplitude-dependent dynamic characteristics derived from perturbation analysis to determine the type and order of nonlinearity and system parameters. For non-parametric identification, the method uses the maximum displacement states to determine the displacement–stiffness curve and the maximum velocity states to determine the velocity-damping curve. Moreover, the SWF method and a synchronous detection method are used for further decomposition of components extracted by HHT to improve the accuracy of parametric and non-parametric estimations. Numerical simulations of several nonlinear systems show that the proposed method can provide accurate parametric and non-parametric identifications of different nonlinear dynamical systems.