Structural assessment and spatial damping identification using low-cost acceleration sensors

A proper characterization of structural damping is one of the most current and challenging topics in mechanics, as damping has multiple sources and available theories are limited in their scope and explanatory power. At the same time, Microelectromechanical systems (MEMS) accelerometers are making inroads in low cost structural health monitoring applications and the available literature on the construction and use of open-source accelerometers is vast with many examples of individual sensor units built with low-cost electronic boards published to date. This paper presents studies on the structural characterization and damage identification of model structures using low-cost wireless accelerometers built with of-the-shelf components. Natural frequencies, damping ratios, damping and stiffness matrices from system output vectors carried out using an output-only modal identification (OMA) and a covariance driven stochastic subspace identification algorithms are presented together with an experimental modal identification (EMA) of viscous and hysteretic damping matrices. These methods are applied to a finite element model and to an experimental model of a 2D steel frame. It is shown that, while the OMA method shows a good accuracy and stability for the FEM data, it can yield unphysical results for the estimation of modal shapes and damping ratios during physical testing and often requires a significant user involvement. On the other hand, the EMA method shows high accuracy and stability and improved levels of automatization. The structure was then altered to simulated progressive states of damage (e.g. beam and column sections were reduced in both models, and bolted connections were loosened). Under these conditions, the natural frequencies dropped between 0.1 % and 22 % due to bolt loosening and between 0.8 % and 25.2% for beam and column section reductions. Overall, this paper shows that the localization of damage is possible from low-cost wireless MEMS accelerometers on the basis of estimated stiffness, viscous and hysteretic damping matrices provided that reasonably low error levels are guaranteed.