Damage Identification via a pseudo-modal approach

The identification of damages in structural systems is nowadays a crucial target. Within this topic, dynamic approaches have been proven reliable and effective and several techniques can be found in the technical literature. In essence, all the proposed identification strategies aim at extracting reliable damage signs from the least amount of response data. On one hand, the modal properties based ones hinge on the dynamic response alteration due to damage and allow for predictive and physically sound interpretations; however, accurate numerical structural modelling, selection of meaningful response signals and uniqueness of the solution of the associated inverse problem are the main source of difficulties encountered. On the other hand, signal processing based techniques allow for direct and accurate tracking of signals changes between undamaged and damaged states in time and frequency; however, physical interpretation of the source of damage responsible for the detected signals changes is often cumbersome.

Nevertheless, real-life structures are liable to worst conditions which may limit the practical application of the latter techniques. First of all, the solvability of the identification problem must be considered, looking at the analytical position of the relevant inverse problem; this issue can be tackled properly choosing the excitation system and the acquisition network.

Our studies propose a new damage identification technique based on the combined use of the decomposition part of the Hilbert-Huang Transform (HHT), known as Empirical Mode Decomposition (EMD), and the experimental modal analysis approach. Therefore the method aims at merging the time-frequency content of the structural response signals provided by the signal processing phase with the dynamic features of the damaged structure provided by the modal perspective. A response based and data-driven pseudo-modal approach for structural damage identification is introduced. The method exploits the Orthogonal Empirical Mode Decomposition to derive multi-frequency sub-signals directly from non-stationary acceleration response. These signals form a complete time-domain basis providing with empirical mode shapes enhancing the modal damage identification capabilities. Then, a damage index comparing undamaged and damaged pseudo-modes is proposed. 

The proposed damage identification technique performance was at first applied to a parabolic double-hinged steel arch with a notch-like damage. Numerical simulations were carried out through an updated model based on extensive experimental results on the undamaged arch. Two selected classical methods, namely RD and COMAC, were found to be insensitive to damage. In particular, whilst RD furnished an erroneous damage location, the COMAC correctly identified the damaged area as long as an extremely accurate model is available. Differently, the technique here proposed clearly indicates the damaged area provided that the orthogonal version of the Empirical Mode Decomposition is adopted and that an energy-based criterion leads the pseudo-mode selection [1,2].

The performance of a pseudo-modal approach has been investigated with respect to the role of the length of the signals and the boundary effects. With reference to a four-storey shear frame structure, the pseudo-modal approach has been applied and the sensitivity of the time-frequency identification technique investigated [3]. Extending the previous studies based on non-stationary acceleration data, strain data are considered in [4]. It is shown that the approach can be effectively applied to different structural response signals. This versatility can exploited in real world applications where different sensor technologies are nowadays widely employed.

In [5] a first experimental validation, which involved a small-scale steel frame, artificially damaged at the third floor with a precision cutting, was reported. Two main contributions turned out from the experimental campaign: first, the noise corruption can be easily avoided looking at fast oscillations with low averaged power; second, an updated energetic selection criterion is introduced to properly tackle variable input energy.

[1] Romeo F., Lofrano E., Paolone A., Damage Identification in a Parabolic Arch via Orthogonal Empirical Mode Decomposition. ASME. Int. Design Eng. Tech. Conf., Volume 8: 26th Conference on Mechanical Vibration and Noise: V008T11A006. doi:10.1115/DETC2014-35529, 2014.

[2] Lofrano E., Paolone A., Romeo F., Damage identification in a parabolic arch through the combined use of modal properties and empirical mode decomposition, Proc. of the 9th Int. Conf. on Structural Dynamics, EURODYN 2014, Porto. A. Cunha, E. Caetano, P. Ribeiro, G. Müller (eds.). ISSN: 2311-9020; ISBN: 978-972-752-165-4, 2014.

[3] Lofrano E., Romeo F., Paolone A., Sensitivity Analysis and Improvement of a Pseudo-Modal Approach for Damage Localization, ASME. Int. Design Eng. Tech. Conf., Volume 8: 27th Conference on Mechanical Vibration and Noise: V008T13A097. doi:10.1115/DETC2015-46715, 2015.

[4] Lofrano E., Romeo F., Paolone A., Damage identification via orthogonal empirical mode decomposition of curvature mode shapes, Proc. of 7th Int. Conf. on Structural Health Monitoring of Intelligent Infrastructure (SHMII 2015), Turin. A. De Stefano editor. ISBN: 9781510821071, 2015.

[5] Lofrano E., Paciacconi A., Paolone A., Romeo F., Experimental validation of a novel pseudo-modal approach for damage detection, Procedia Engineering, vol. 199, pp 1943-1948, 2017.