Exploring Adaptive Data Analysis for Non-parametric Identification of Linear and Nonlinear Structures with Time-varying Properties
Riferimenti GIAN PAOLO CIMELLARO
Descrizione Fundamental to the widely used Hilbert-Huang Transform, the mathematically-unproven empirical mode decomposition (EMD) faces significant challenges in dealing with noisy data, weak intermittent fluctuations, closely-spaced modes of vibration, and temporal changes in the mode of vibration. This proposal seeks funds to explore an alternative to the EMD, analytical mode decomposition (AMD) in the context of a unified theory (analogous to modal analysis in Structural Dynamics) for non-parametric identification and damage detection of time-varying systems (both linear and nonlinear structures). The main idea behind the AMD and Hilbert Spectral Analysis (HSA) is to accurately decompose a non-stationary or nonlinear time response of a structure into many components whose frequency characteristics can be completely separated by predetermined time-varying bisecting frequencies. As the divided frequency ranges become sufficiently narrow, a small perturbation of structural properties will be amplified in some of the narrowband components. As such, instantaneous system properties can be identified with high accuracy and high sensitivity to the perturbation from the “zoom-in” components, and their difference from a corresponding undamaged structure can be used for damage detection.
To date, the AMD-HSA has been demonstrated to be a powerful adaptive data analysis method with frequency-modulated analytical signals and simulated structural responses where bisecting frequencies are either constant or can be expressed into a time function. To explore its full potential, the AMD-HSA must be further studied and validated with laboratory and field test data of engineering structures and its key element, time-varying bisecting frequencies, must be evaluated in a systematical approach.
The objectives of this study are: 1) to propose, develop and characterize adaptive wavelet transform (AWT) for the determination of time-dependent frequencies so that uniform and high frequency resolution can be achieved near the ridge lines of wavelet transform scalograms; 2) to establish a new instantaneous frequency relationship between a time-varying structure and its response, develop a computationally efficient solution procedure for the instantaneous frequency, and quantify its advantages over existing relationships in the literature; and 3) to study robustness, accuracy, and efficiency of the proposed AMD-HSA method and compare it with existing system identification techniques using both laboratory and field test data sets obtained from complex structures with various nonlinear mechanisms.
Conoscenze richieste English speaking
Scadenza validita proposta 26/04/2020 PROPONI LA TUA CANDIDATURA