Identification of structural damage based on locally perturbed dynamic equilibrium : a "pseudo-excitation" (PE) technique

Abstract

In recognition of the obvious limitations of most global vibration-based and local guided-wave-based damage detection techniques, a novel inverse damage identification approach, named the pseudo-excitation (PE) technique in this study, was developed in this PhD thesis project, by canvassing the local perturbance to equilibrium characteristics of a structural component under inspection. Characterized by higher-order spatial derivatives, this approach has proven sensitivity to structural damage such as interfacial debonding. Most importantly, it requires neither benchmark structures nor baseline signals, neither global models nor additional excitation sources, as long as the structure undergoes steady vibration under its normal operation. Independent of a global model, prior knowledge of structural boundary conditions is not a prerequisite. Numerical and experimental validations were performed first on beam-like structures, and then were expanded to the two-dimensional domain, showing the capacity of the method to characterize multiple damage in a plane structure comprising both beam and plate components. In addition, the method manifests its effectiveness in detecting interfacial debonding between different structural components by reconstructing the distribution of interfacial forces. It was also recognized that the interference from measurement noise was a major hurdle for this method from achieving satisfactory detection results, because of the higher-order derivatives involved in the constructed damage index (DI). That issue led to the development of various de-noising strategies in this thesis: low-pass wavenumber filtering (LWF) was first proposed based on spectrographic analysis whereby the signal components of DI in higher wavenumber domains were screened out by designing a low-pass filter; the adjustment of measurement density (AMD) was developed as an alternative for noise reduction. It was found that by optimally maneuvering the ratio of the distance between two adjacent measurement points to the wavelength of vibration, measurement noise could be considerably suppressed but at the cost of sacrificing a certain amount of the computation accuracy of the finite difference. Owing to the adaptability of the method and the flexibility of the de-noising techniques, data fusion algorithms could be adopted additionally using signals captured from a very limited number of tests, further enhancing the robustness of the method in noisy conditions. In principle, the PE method is applicable to a complex system comprising various structural components, provided that the local equilibrium relationships of the components are known a priori. As an application, a steel-reinforced concrete slab with multiple debonding regions was inspected experimentally using the developed technique, revealing the potential of the method to be used in damage detection in real civil engineering structures. Subsequently, a "weak" formulation of the PE technique was proposed as an independent implementation to further enhance the noise immunity of the approach. By introducing weight functions in integration computation, the weak formulation gives rise to a variety of possible forms, whereby different damage indices associated with different forms can be constructed, facilitating the extraction of damage-related features. As two examples, Continuous Gauss Smoothing (CGS) and Virtual Vibration Deflection (VVD) were introduced. The former relied on the original weak formulation with the weight function set as the typical Gaussian function, while the latter was base on a modified version of the weak formulation, providing a new approach to vibration-based detection utilizing densely measured vibration displacements and sparsely measured dynamic strains. The effectiveness of the CGS was verified experimentally using an aluminum beam-like structure containing multi-cracks; and numerical studies were conducted to validate the effectiveness of VVD.