Using multiple nondestructive test data types and data sets for condition assessment of bridge decks

This research examines the potential of structural parameter estimation for the purpose of condition assessment of existing structures. Structural parameter estimation is the art of reconciling an a-priori analytical finite element model of the structure with test data. The purpose of this research is to use non-destructive test (NDT) data to identify stiffness parameters of a laboratory grid representing a bridge deck created and tested at the University of Cincinnati Infrastructure Institute (UCII). Measured static displacements, static strains and modal data sets were used for parameter estimation. Prior to using the NDT data, several techniques were developed as unique contributions to parameter estimation to address a number of technical issues associated with parameter estimation.; The new parameter estimation formulations, stacking and combining, which are unique contributions of this research, allow for separate load cases and error functions to be used concurrently for parameter estimation with individual subsets of measurement locations. To facilitate the combination of error functions, error function normalization is also presented as a unique contribution to parameter estimation. These techniques improve performance of the parameter estimation algorithms for both simulated scenarios contaminated with measurement error and scenarios using NDT for condition assessment.; Several parameter estimation scenarios using a finite element model of the UCII-Grid were successfully completed using error function normalization, stacking and/or combing. The UCII-Grid stiffness parameters were updated to reflect the non-destructive test data. Only scenarios that used error function normalization and stacking were successful, in terms of physical meaning for the parameter estimates. The unsuccessful cases showed that no amount of numerical manipulation can make an ill-conditioned scenario produced meaningful results.; Two new error functions, the modal eigenvalue-based and eigenvector-based error function, were developed and implemented for parameter estimation. These error functions represent the residual eigenvalues and eigenvectors, respectively measured at individual subsets of degrees of freedom of the finite element model. Both the eigenvalue-based and eigenvector-based error functions were successful in performing parameter estimation for a frame model. However, these error functions were not used with the NDT for the UCII-Grid but they do provide the user more freedom when performing parameter estimation.