| Robust control techniques have enabled engineers to create uncertain models which are able to describe any differences between the model and experimental system with uncertainties defined as a combination of exogenous inputs and plant perturbations. Subsequently, robust model validation techniques arose to provide a guarantee that the uncertain model is able to recreate all observed experimental data. As a result, the complete model set is robust to any model inaccuracies or external noise. At the same time, the technique of model-based identification was developed in the robust control framework to identify the dynamics resulting from unmodeled or under-modeled components in mechanical systems. The approach controls the nominal model in order to minimize the error between its response and that of the experimentally identified system. The resulting controller estimates the difference in dynamics between the model and actual system, also known as the unmodeled dynamics.;In this work, a damage identification technique is developed which combines model validation and model-based identification for robust control relevant structural health monitoring. The method will both detect the presence of damage and identify the local change in dynamics due to the damage in a robust control framework. As a result, the damage detection will be robust to mismodeling and noise. Additionally, the identified damage dynamics will be defined with an uncertainty bound which will serve the dual purpose of a definition for robust control and a quality estimation of the nominal damage dynamics. The new technique is demonstrated experimentally on a rotordynamic test rig. First, feasibility of the method is verified by the identification of a fully-open seeded crack in a non-rotating shaft. Finally, the precision of the method is demonstrated through identification of a breathing crack in a rotating shaft. |
