A realtime identification algorithm for active control of smart structures

During the last years, more and more mechanical applications saw the introduction of active control strategies. In detail, vibration suppression is often associated to the need of increasing system performance or improving the structure health and lifetime. This goal can be achieved considering both passive and active solutions. Among the active ones, many different control strategies have been proposed and developed, such as modal and resonant control, optimal control, robust control, etc. For these techniques, the knowledge of the system dynamic behavior is very useful to tune the control parameters so that the best performance can be achieved. This can be done through a numerical model of the structure or through an experimental identification. Anyway, if a non-linear and/or time variant system is considered, an online identification of the system dynamics is a key point to keep the control effective over time. For this reason, this paper proposes a real-time algorithm, based on ARMAX models, which continuously estimates the modal parameters of a structure. ARMAX models are based on a certain number of poles and zeros but, as known in literature, it is necessary to fictitiously increase the model order to achieve a correct identification. To overcome this limitation, the paper proposes to couple the ARMAX formulation with a Recursive Subspace Tracking algorithm, identifying the main components of the measurement signal and provides them to a number of reduced order ARMAX models. The paper also includes a numerical and experimental investigation of the proposed algorithm, performed on a carbon-fiber plate with piezoelectric sensors and actuators.