Damage Localization for Structural Health Monitoring: An Exploration of Three New Vibration-based Schemes

TLDR: Three new vibration-based damage localization schemes are proposed that, in one way or another, address the noted sensitivity issue and are considered the two main contributions of this thesis.

Abstract: The tendency within engineering is to build increasingly large structures while minimizing material use. Needless to say, this leads to less conservative designs and, as such, an increased demand on regular inspection of the structures to ensure that they maintain adequate reliability through their life cycle. The inspections are conventionally conducted by sending out trained personnel to visually assess the integrity of the structures. This procedure can, however, be associated with high costs due to operational downtime and, for structures located in terrain not easily accessible, transportation. A general consensus is that vibration-based structural health monitoring (SHM), which involves implementing a damage identi cation strategy to monitor structural integrity using vibration measurements, can play a role in reducing the inspection costs. Numerous SHM techniques have been suggested, and while the task of detecting whether damage is present or not has been resolved with reasonable success, a reliable solution has not yet been presented for the next logical step; namely, to locate the detected damage. There are many reasons as to why vibration-based damage localization has not yet found the level of industrial applicability that one would anticipate after decades of research. One of these reasons is undoubtedly that the vibration features used in the process, such as modal parameters, lack sensitivity to damage compared to the sensitivity to noise and other variabilities. In the present thesis, three new vibration-based damage localization schemes are proposed that, in one way or another, address the noted sensitivity issue. The rst exploration is the CWT-GDTKEO scheme, whose methodological premise is to seek for damage-induced changes in signal-processed mode shapes of the structure in question. More speci cally, the scheme incorporates continuous wavelet transformation (CWT) and a generalized discrete Teager-Kaiser energy operator (GDTKEO) to capture these changes, and the damage location is attained using a simple metric comparing processed signals from the states prior and posterior to damage. In this way, the scheme relies on su ciently accurate estimation of the required mode shapes, which, in many application scenarios, can be di cult to achieve due to noise and/or poor excitation. The obvious drawbacks of the CWT-GDTKEO scheme have led to exploration of what are considered the two main contributions of this thesis. The rst one is the Subspace Exclusion Zone (SEZ) scheme, which, under certain input conditions, circumvents system identi cation. The scheme locates damage by reconstructing shifts in measured eld quantities using subspaces indexed by postulated boundaries, the so-called exclusion zones (EZs). The methodological concept rests on the fact that shifts in any eld quantity outside the boundary of an EZ encompassing the damage can be generated from stress elds acting on the aforementioned boundary. The EZs, which are formed in a theoretical model of the structure prior to damage, are of user-de ned size, thus information on size and type of damage is precluded to provide a net robustness gain. Application examples are presented that clearly demonstrate the robustness of the SEZ scheme in instances allowing for a system identi cation-free con guration. The second main contribution of the thesis is the Shaped Damage Locating Input Distribution (SDLID) scheme, which operates unconditionally free of system identi cation. The methodological premise is to deploy controllable inputs that are tailored to actively interrogate one structural subdomain at a time. When the subdomain containing damage is rendered dormant, the e ect of damage and, as such, its induced shift in steady-state vibrations are canceled. In this way, the SDLID scheme facilitates damage localization using only few output sensors; in fact, one well-placed may su ce. This low demand on output sensors is an attractive feature, which is conventionally only achieved when employing an approach based on guided waves. However, unlike this approach with high-frequency waves, which is merely suitable for local integrity inspection because of small wavelengths and high damping, the SDLID scheme can operate in a broad band of frequencies.