A bespoke signal processing algorithm for operational modal testing of post-tensioned steel and concrete beams

The extraction of modal properties, specifically natural frequency, damping ratio and mode shape is a difficult task, especially when output-only data is measured. The accuracy of the estimation these modal properties is compromised by noisy signals, and signal filtering is required to suppress unwanted frequency content. Care is required however to avoid over-filtering of the output data, which can eliminate valid structural frequency content if required care is not exercised. This paper describes the development of a bespoke signal processing algorithm to extract the modal properties of both simply supported post-tensioned steel and concrete sections. Dynamic impact testing was conducted on a series of different post-tensioned steel rectangular hollow sections, and 9 different post-tensioned concrete beams, each with differing straight profiled post-tensioning strand eccentricities. Acceleration time-history data was recorded for each of the steel and concrete beams via an accelerometer. This data was subsequently processed, first centring the acceleration-time history using a moving average filter, and subsequently removing any zero drift in the accelerometer via a second order low pass Butterworth filter. Electrical noise was then removed via a notch filter. The accelerometer data was then smoothed in the time domain. The Fast Fourier Transform (FFT) was applied to the signal to convert into the frequency domain and finally a bespoke peak-picking algorithm was invoked to extract the natural frequencies of the beams. A comparison is subsequently made between the accuracy of the estimation of the modal properties of the steel and concrete beams for filtered and unfiltered data, and a sensitivity analysis of the filtering and peak picking parameters is conducted to determine the effect that this has on the accuracy of the estimation of the modal parameters. The results show the effectiveness of the bespoke signal processing algorithm in increasing the accuracy of the estimation of the modal properties as opposed to the raw unprocessed signals.