Development and application of parametric system identification algorithms for structural health monitoring

Over the past years, there has been an increasing interest in the efficient simulation and identification of structural system behavior. The subject of structural health monitoring has become even more critical due to the obvious need for maintaining the deteriorating infrastructure. The latest advances in sensor technology have allowed for an effective and low cost instrumentation and monitoring of civil structures by deploying relatively dense sensor arrays. This allows for the detection of unusual structural behavior, the indication of the approximate position of a problem, and the triggering of remedial actions.;This dissertation, begins by exploring the application of Bayesian based techniques for the identification of non-linear systems. The non-linearity in structural response usually stems from hysteretic behavior. Existing techniques are evaluated and compared on the identification of the governing parameters of a system and novel algorithms are introduced for the improvement of the estimation on the class of problems we are dealing with herein. The importance of fusing data from heterogeneous, non-collocated sensor readings is investigated. Additionally, an experimental application is presented for the identification of a non-linear hysteretic system where not only the system parameters are unknown but also the nature of the analytical model describing the system is not clearly established. To this end a parallel identification scheme employing candidate models of varying complexity is introduced. Finally, we study an identification scheme combining Genetic Algorithms and the Extended Finite Element Method for the problem of non destructive sensing (localization of faults on plates based on strain sensor readings). A novel GA mutation scheme is introduced for the improvement of the GA convergence rate and the validity of the proposed method is tested experimentally.