Improving Substructure Identification Using Structural Control with Limited Sensor Measurements

Substructure identification is a structural health monitoring technique that identifies a reduced-order model of structural behavior using local vibration measurements. The authors’ prior research on this inherently decentralized technique demonstrated that substructure identification can detect and localize stiffness changes in a shear building. However, identification performance varies from story to story, with parameters of a few stories not satisfactorily identified. To overcome this limitation, the authors and colleagues showed that a structural control device can be used to temporarily change the dynamics of the structure to improve identification performance; however, prior studies assumed full-state feedback, which is not achievable in full-scale implementations. This paper investigates structural control using limited sensor measurements to improve substructure identification. The substructure method is first summarized, followed by an error analysis that can predict a priori the level of error in story parameter estimates. Then, a novel controlled substructure identification procedure is introduced with constraints to ensure that the controlled responses improve substructure identification while reducing overall structural response. State feedback is utilized, with a Kalman filter to estimate states using various sets of acceleration sensor measurements to understand the performance trade-offs. The implementation methodology using an object-oriented programing paradigm is explained; a 3.66-m (12-ft) four-story single-bay steel structure model, subject to low levels of ground motion, is the test bed used for numerical simulations to demonstrate the proposed methodology. The control strategies that include a sensor measuring ground acceleration are shown to be nearly as effective in estimating stiffnesses as the full-state feedback control and all controlled strategies perform better than the uncontrolled case.