Anomaly identification in mechanical structures exploiting the inverse finite element method (ECCM -ECFD 2018 conference)

One main limitation to the implementation of Structural Health Monitoring (SHM) systems in real structures is the influence of different boundary conditions with respect to those adopted during SHM system design, especially for what concerns the loads, potentially leading to damage misclassifications. In this context, the inverse Finite Element Method (iFEM), recently developed for shape sensing of shell structures, can be used to reconstruct the displacement field, thus the strain field, everywhere in a component on the basis of just few strain sensors placed in discrete positions and without requiring any a-priori knowledge of loads or material properties. This work proposes a methodology to perform SHM exploiting the iFEM algorithm for strain reconstruction. In particular, an anomaly index is defined based upon the comparison between the strain read at a target sensor location and the one reconstructed, in the same position, through the iFEM algorithm. When the analyzed structure is in a “healthy” condition, the two values match, otherwise they do not. The defined anomaly index enables to identify both the presence and the position of a defect within the structure without being dependent on the modelled boundary load condition. Computation efficiency is ensured by the iFEM algorithm itself. A very fast reconstruction of the component strain field is achieved once a sensors grid is established within the structure, meaning the method can be easily implemented in an online monitoring system. Though the method formulation is general for an arbitrary component geometry and damage type, the proposed methodology is experimentally tested by means of a clamped plate subjected to fatigue crack propagation. The results underline the method attractiveness for its ability to correctly detect both the presence and the location of the damage without any prior information on the boundary load condition and with a low computational effort.