Data Sampling Frequency Impact on Automatic Operational Modal Analysis Application on Long-Span Bridges

Bridge monitoring projects based on vibration data analysis tend, since the early 2010s, to process vibration acceleration data using a time-domain system identification method use an automatic operational modal analysis (AOMA) algorithm to extract the modal properties, without the involvement of an operator. This work highlights the impact that the acceleration data sampling frequency has on the outcome of some of the AOMA algorithms. Acceleration datasets, downsampled to different frequencies, from the Hardanger Bridge are processed by the Magalhaes 2009, Neu 2017, and Kvåle 2020 AOMA algorithms. It is shown that that the best results in terms of modal detection rates and number of errors are obtained for sampling frequencies between 10 and 20 Hz. Additionally, no algorithm is more impacted than another by the different sampling frequencies.