A Polyreference Least Square Complex Frequency domain based statistical test for damage detection

Abstract

Some current approaches in damage detection have led to the implementation of statistical algorithms, based on the so-called “local approach” assumption. The reference parameter is the modal signature of a state-space model of the studied system, and the tests aim at detecting small deviations in this signature without explicitly computing it. Moreover, the system inputs are assumed to be random white noises, modeling all the various and unknown influences on the system, either environmental or operational. Working with the outputs in time-domain, various tests were designed and experimented - e.g. the subspace-based chi-square test. On the other hand, efficient system identification algorithms working in the frequency domain have been designed, first dealing with experimental conditions - meaning there are known controlled inputs - but also offering promising perspectives with regards to the operational conditions - without any measurement of the inputs. In this paper, a frequency-domain statistical test for change detection is proposed, based on this recent frequency-domain modal analysis method and on the local approach to change detection.