Experimental and analytical studies on pedestrian induced footbridge vibrations

This paper presents results from experimental study on human-induced vibrations of three lively footbridges in Reykjavik. The project was funded by the Icelandic Public Roads Administration with two main focus areas; validating the FE-models used at the design stage in terms of dynamic characteristics and to measure the response of the bridges to human induced excitation such as walking, running and jumping.

Two of the bridges were a part of a design competition related to the realignment of an existing highway in the city centre of Reykjavik in 2005. The bridges have similar cross sections and overall appearance, but different lengths and layout. The first bridge is a continuous 86 m long post-tensioned concrete bridge in five spans (max. 23,5m), the second bridge is 169 m continuous post-tensioned concrete bridge in eight spans (max. span 27,1m).

The third bridge is a 54 m long steel bridge with two equal spans crossing the same highway and was built in 2000. A commercially available finite element program (SAP2000) was used in the design phase to model the bridges. The FE-models were updated after the initial tests in order to have the frequencies and damping of the fundamental vibration modes corresponding to the measured values.

The models were subsequently used to calculate the predicted acceleration according to the preliminary version of the Eurocode (ENV 1992-2: Concrete bridges) using time-history analysis with a moving load as representative for a single pedestrian. The load models describing human-induced vibrations on structures in current literature and standards are explained, both for a single person walking or running and crowd loading.

The measured vertical acceleration induced by single pedestrians was compared against the predictions and it was found that all the bridges satisfy the criteria set in this standard, i.e. both the calculated and the measured response from single pedestrians were within the acceleration limits. However, as the load model used in the simplified code procedure does not represent the actual load induced by the pedestrian at different pacing rates large differences were observed between the measured acceleration response and the one predicted by the code.

Instead, a slightly improved simplified load model using frequency dependent load factors gives better results.