| Structural health and safety play a major role in all facets of human daily lives. Over the past few decades significant advancements have been made in structural damage detection and health management in a wide range of engineering disciplines and practices, including but not limited to aerospace, power generating plants, infrastructure systems, and manufacturing. Two main thrust areas of research in this field include the development of methodologies/algorithms for detection of damage and/or changes in the dynamic characteristics of the system, and the sensing/detection devices for capturing the required data/information. A third and evolving area is the integration of these two thrusts and the development of integrated systems that can "manage", and "adapt" in an "intelligent" sense, the subsequent actions that need to take place in order to maintain the integrity of the system subject to the external environment and/or loading conditions. However, majority of the developed techniques fail to take into account the important effects of uncertainty presented in sensing, system modeling, and material behavior associated with dynamic systems. These uncertainties could greatly affect the structural performance and health management, which leads to challenging issues such as reliability and life prediction of the structure. In order to address these important factors, the application of the probabilistic and reliability analysis techniques to structural health management has emerged as an active research area in recent years.; Bayesian probabilistic analysis is such a technique in which the uncertainties could be related with a mathematical model---probability distribution function---which interprets the measurement of confidence interval. The posterior probability distribution is known as an expression for the statistical knowledge of a system after a set of measurements is made. The Bayesian approach is a powerful way to continuously optimize the "posterior" probably density function (pdf) by adapting the predefined "priori" pdf based on "new" measurements. On the basis of Bayesian analysis, it is shown to be possible to perform statistical based system identification, structural damage detection and reliability assessment, as part of structural health management.; In the first stage of this thesis work, a Bayesian based system identification approach was developed to identify system parameters provided so that inherent uncertainties and probabilities of system changes and/or environmental disturbances are taken into account. It is obvious that the nature of changes encountered the system models is critical to monitoring and managing the integrity of the structural systems. In this part of the work, system changes were modeled as random variables with certain statistical properties. The effects of priori definition and different data sampling techniques were studied. To explore the application of this Bayesian based system identification approach to structural health management, the probability density function (pdf) profiles of model parameters were studied to quantify the uncertainties associated with the estimated parameters. By analyzing the posterior pdf inference, the reliability parameters of interest could also be obtained from the available data.; Structural health monitoring, damage detection and structural reliability are usually considered as the sequential components in a structural health management chain. It is the ultimate goal of structural health management to achieve a significant improvement of the structural reliability. Therefore, the second stage of this thesis work was devoted to the study of system reliability. A reliability analysis package developed in MATLAB-PROBES was enhanced with its functionality in this work. The enhancements include the new capabilities of performing analysis to correlated, non-normally distributed random variables and the added functionality to obtain the statistical informati... |
