Misclassification error rejection in structural models for spatial architectures

Vibration mode pairing via modal coupling and modal assurance criteria is a task encountered in different applications of Structural Health Monitoring. Above these, Sensitivity Analysis, and Model Updating need a correct estimation of the pairing between a reference set of vibration modes (that can be both numerical or experimental) and different sets of numerical vibration modes, generated by the variations of structural model parameters. The incorrect pairing, or coupling, results in error propagation during the sensitivity analysis or model updating, generating biased results. In this paper, the authors propose a reliable and efficient method to reduce the effects of mode pairing (i.e., mode coupling) errors in the calculation of the sensitivity of modal data to mechanical parameter variations, even under complex conditions such as complex numerical models of spatial structures, where the vibration modes are characterized by, for example, high-order mode shapes and high probability of confusing correlated mode shapes. The proposed method is based on the conjunction application of Hilbert-Huang Transform to remove outliers, and Gaussian Mixture Models to alleviate the elimination of meaningful information during the outliers rejection phase. In particular, knowing the actual sensitivity of the vibration modes not only allows to optimize the subsequent automatic model updating phase, but even before that, it allows to optimize the design of a permanent monitoring system, for example by predicting the direction and position of the sensors which maximize the extractable modal information considering that the mechanical parameters are continuously subjected to Environmental and Operational Variations, or again, would allow the optimization of extended experimental campaigns, suggesting additional number of mechanical tests for those structural components which most influence the modal behavior of the structure.