An experimental investigation of change detection in uncertain chain-like systems

Promising ongoing research on “smart” sensing technologies is offering low-cost alternatives and new opportunities for large-scale SHM. Networks of sensors with wireless communication and computational capabilities can be used to increase the spatial resolution of data collection while providing a distributed computing framework for implementing structural health monitoring algorithms. Robust and practical SHM methodologies being able to rapidly and accurately detect and assess changes in the monitored system are required to be at the core of these “smart” structures. A data-driven non-parametric identification technique is used to implement a robust change detection methodology for uncertain MDOF chain-like systems that can be implemented in densely distributed smart-sensor networks. Experimental data from a test-bed structure tested at Los Alamos National Laboratory are used to evaluate the effectiveness and reliability of the proposed SHM methodology. The results of this study showed that the proposed approach was able, in a rigorous statistical framework, to confidently detect the presence of structural changes, accurately locate the structural section where the change occurred, and provide an accurate estimate of the actual level of “change”. Additionally, a full-order finite element model of the test structure, as well as the results from the experimental modal identification using the ERA algorithm were employed to validate the results obtained in this change-detection study.