| In contrast to conventional non-destructive evaluation methods, dynamics-based damage detection methods are capable of rapid integrity evaluation of large structures and have received considerable attention from aerospace, mechanical, and civil engineering communities in recent years. However, the identifiable damage size using dynamics-based methods is determined by the number of sensors used, level of measurement noise, accuracy of structural models, and signal processing techniques. In this thesis we study dynamics of structures with damage and then derive and experimentally verify new model-independent structural damage detection methods that can locate small damage to structures.; To find sensitive damage detection parameters we develop a higher-order beam element that enforces the continuity of displacements, slopes, bending moments, and shear forces at all nodes, and a higher-order rectangular plate element that enforces the continuity of displacements, slopes, and bending and twisting moments at all nodes. These two elements are used to study the dynamics of beams and plates. Results show that high-order spatial derivatives of high-frequency modes are important sensitive parameters that can locate small structural damage. Unfortunately the most powerful and popular structural modeling technique, the finite element method, is not accurate in predicting high-frequency responses. Hence, a model-independent method using dynamic responses obtained from high density measurements is concluded to be the best approach.; To increase measurement density and reduce noise a Polytec PI PSV-200 scanning laser vibrometer is used to provide non-contact, dense, and accurate measurements of structural vibration velocities. To avoid the use of structural models and to extract sensitive detection parameters from experimental data, a brand-new structural damage detection method named BED (Boundary-Effect Detection) is developed for pinpointing damage locations using Operational Deflection Shapes (ODSs) measured by the scanning laser vibrometer. The BED method decomposes an ODS into central and boundary solutions by using a sliding-window least-squares fitting technique. Because boundary solutions are non-zero only at damage sites as well as structural boundaries, they can reveal damage locations. Three signal processing methods are developed for use in the BED method: (1) processing 1-D ODSs (i.e., beams) using a sliding-window curve-fitting technique, (2) processing 2-D ODSs (i.e., plates) using a sliding-window surface-fitting technique, and (3) processing 2-D ODSs using a sliding-window curve-fitting technique.; Experiments are performed on isotropic beams and plates with different types of damage. At a damage location on a beam, results show that the boundary solution of slope changes sign, the boundary solution of displacement peaks up or dimples down, the fitting error peaks up, and second and third spatial derivatives show sudden changes. For plates, results show that sensitive damage detection parameters are boundary solutions and parameters related to bending moments and shear forces because they show significant sign changes or peaks at damage locations. All the results show that the BED method is more sensitive than other dynamics-based methods using curvatures and/or strain energies. Moreover, because the BED method works without using any structural models or historical data and it provides multiple damage detection parameters for users to ensure the identified damage locations, it is a reliable method and is applicable to actual structures. |
