Abstract

An increasingly used approach for health monitoring is based on error localization techniques. They are applied on an analytical model of the nominally healthy structure and produce local values of discrepancies between the model and the measurements. In this way, damaged regions are identified, and the selection of the model parameters (to be updated) becomes easier. This helps the optimization procedure to reconciliate the model and the measurements, since the design space is reduced. Among the different techniques used to indicate regions where the analytical model presents parameter errors, one finds those that are based on output comparison: COMAC, MAC variation. These methods quantify the loose of correlation between paired modes associated to a degree of freedom. They assume that local model errors produce local changes in the modes (which is a case dependent assumption). They present the advantage of not requiring the use of expansion or reduction techniques but they are limited by the mode pairing process and by the norms used for the comparison. Another approach is to use the model equilibrium conditions. In this category, one finds the force residual technique and the method based on the minimization of errors on constitutive equations (MECE). A drawback of these techniques is that they require a matching process a priori. The aim of this paper is to analyze the limitations of this last class of techniques to locate model errors.