Benchmark study on system identification through ambient vibration measurements

There are several ambient vibration system identification techniques available that were developed by different investigators or for different uses. Consequently, the question occurs to the comparison among those analysis techniques. The benchmark study of the paper is intended to compare system identification techniques for evaluating the dynamic characteristics of a building from ambient vibration data. The averaged normalised power spectral density method (also called peak picking method) and the stochastic subspace identification method based on the singular value decomposition are used in the study. A 15 storey reinforced concrete shear core building has been chosen as a case for verification. The results have shown that both techniques can identify the eigenfrequencies and the mode shapes. Damping ratios can only be determined by the stochastic subspace method. The stochastic subspace identification technique can detect frequencies that are possibly missed with the peak picking method and gives a more reasonable modal shape in most cases. However, the stochastic subspace technique is more time consuming. For real applications, it is suggested that the peak picking technique could be used on site to judge the overall dynamic characteristics of the structure. And then, the stochastic subspace identification technique could be applied afterwards to detail or to ensure the results.