Adaptive POD-based reduced order modeling and identification of nonlinear structural systems

The health monitoring of structures and infrastructures exposed to aging and/or extreme loading conditions is increasingly recognized as a timely issue, not only in civil engineering. To move towards real-time structural health monitoring (SHM) procedures, the adoption of reduced-order models (ROMs) obtained through proper orthogonal decomposition (POD) was proposed in [4]. In the case of a changing structural health, the ROM would need to be continuously re-trained so as to prevent the results to be affected by biases. In [2] a procedure was offered to avoid such time consuming re-training stages; in this work, we further discuss the proposed framework. The chance to attain even larger speedups through a coupling of such reduced-order modeling procedure and adaptive domain decomposition strategies, see e.g. [3], will be also commented during the Conference.