Model Updating of a 3D Frame Structure through Minimization of Modal Dynamic Residual

For higher simulation accuracy, finite element models of a structure can be updated based on modal analysis results from field testing. Selected structural parameters are updated by solving an optimization problem. The majority of model updating studies attempt to minimize the difference between experimental and simulated modal properties, such as resonance frequencies, mode shapes, etc. This category is referred as modal property difference approach. However, the approach minimizing modal property difference usually suffers from inaccuracies, which is caused by that fact that the minimization objective function (comprising of modal properties differences) is highly nonlinear or non-sensitive to physical parameters being updated. Another category of model updating methods will be referred as modal dynamic residual approach, which minimizes modal dynamic residual from the generalized eigenvalue equation involving stiffness and mass matrices. Early studies have shown that the modal dynamic residual can be a more sensitive objective function for model updating. Furthermore, an iterative convex optimization procedure has been proposed to efficiently solve the optimization problem. This research performs a comparative study between two model updating approaches (minimizing modal property difference or modal dynamic residual). A space frame structure is adopted for the study. Numerical simulation is first performed for validating both approaches. For further validation, Field testing is conducted on a space frame pedestrian bridge on Georgia Tech campus using wireless sensors. The bridge structure is divided into four substructures. Dense wireless sensors are deployed on one substructure at a time, at high spatial resolution. Using the wireless sensor data, modal characteristics for each substructure are extracted, and then assembled to obtain the global modal characteristics.  The model updating results using both approaches are compared. The approach minimizing modal dynamic residual shows certain advantages in computational efficiency and accuracy.