Dynamic Testing and Modeling of a Superelevated Skewed Highway Bridge

Created in response to the aging infrastructure in the United States, the Long Term Bridge Performance Program (LTBPP) under the Federal Highway Administration (FHWA) proposes to assess the long-term performance of representative bridges through nondestructive evaluation (NDE) techniques and visual inspection. For consistency, a set of guidelines is needed to define the procedures for testing each bridge. The NDE techniques involve dynamic testing, and the protocol for this testing has yet to be finalized.

To evaluate the dynamic testing guidelines, a 103 ft single-span, simply supported highway bridge was dynamically tested. The test bridge was characterized by a skew of 34° and superelevation around 4%. Forced vibration testing involved an impact hammer with accelerometers measuring the response. Resonant frequencies were identified from the data by picking peaks from the magnitudes of the frequency response functions (FRF). Eleven modes were identified with frequencies ranging from 2.75 Hz to 22.5 Hz. Mode shapes associated with each mode were constructed using the imaginary components of the FRFs. The half-power bandwidth method was used to estimate the damping for each mode, with values ranging from 1% to 5% of critical damping.

Finite element (FE) models of the bridge were constructed in the commercial FE software Abaqus. The effects of adding and removing superelevation and skew, varying mesh refinement, and changing boundary conditions on modal parameters were thoroughly investigated. FE models were compared to the experimental results by directly comparing frequencies and using the modal assurance criterion to compare mode shapes. Support conditions of the actual structure were bounded using the results of the comparison.

Much insight was gained about forced vibration testing as applied to a full-scale bridge. The spectral resolution of the data proved to limit the accuracy and confidence of detecting closely-spaced modes and calculating damping estimates. Also, a more controlled method of exciting the structure was desired, such as using a shaker with a known input. Resonant frequencies of the FE models were sensitive to changes in boundary conditions, with some frequencies doubling. Both changes in boundary conditions and including skew and superelevation noticeably affected the mode shapes. When compared to the experimental results, the models with idealized roller and pin boundary conditions provided the best correlations based on resonant frequencies and mode shapes.