Short-time record analysis for condition monitoring of bridge stay cables

Cable-stayed bridges have emerged as the preferred design for new intermediate and long span bridges. However, ambient excitation of large-amplitude vibration response of the cables has led to failures requiring additional maintenance efforts and a need for better approaches for condition monitoring. In this paper, a novel signal processing approach is applied to automate the data analysis for determining fundamental frequencies of the stay cables as an indication of condition. With the large number of cables typical on each bridge, this process must accurately represent the cable response and determine the first mode natural frequency while being efficient as possible. This is accomplished by using short acceleration time histories of cable response to ambient traffic, approximately one minute in length. This approach is in contrast to the approach generally associated with low-frequency bridge deck field testing that uses longer time records, more than fifteen minutes in length. Automated analysis of short time records is possible due to the unique nature of cable response and is accomplished by applying a second Fourier transform to the initial spectra of the time histories. This "spectrum of a spectrum" is referred to as the "cepstrum." In essence, this analysis finds the frequency of the harmoniCS occurring in the spectrum of the time history. Bridge stay cables, like rotating machinery where this approach is typically used, exhibit response spectra that include many harmonic peaks. The dominant cepstrum result is the period of the fundamental frequency, from which the fundamental frequency can be determined. In this paper, the automated analysis approach is presented in detail, along with evaluation of accuracy, resolution and consistency of results. Data records over several seasons and after several years from over four hundred cables of four bridges have been analyzed with this approach. Selected results are presented using the approach, including a comparison to long time histories recorded from stay cables. Comments on usefulness for condition monitoring are included.