Robust detection of changes with application to structural systems

The objective of this paper is to present some fault detection/diagnosis schemes, known for their robustness concerning the system undermodelling and variations in experimental conditions. The first approach, in frequency domain, is based on the results of high-order black-box identification theory, and the second makes use of an impulse invariant transformation to obtain the equivalent system having as input a stationary uncorrelated sequence, and as output the corresponding response; this last approach reduces the problem of change detection in systems to one of change detection in signals. These approaches were applied for the analysis of the dynamic behaviour of a vibrating structure, a multistory concrete building, subject to a transient natural excitation.