Non-biased estimation of modal parameters of multi-component ambient vibrations in high-rise buildings using a self-governed random decrement technique

摘要

The estimation of modal parameters of a structure by applying the random decrement technique has been introduced since late 1960's, and is widely used because of its simplicity and rapidity. Since this technique can be applied to mono-component signals only, a filtering process is a necessary preliminary step for getting the random decrement signature when applying on multi-component signals. But filtering is the most critical step with a high influence on the accuracy of the modal parameter estimation, especially when the signal is composed of very closely-spaced modes. In this paper, we propose to apply the random decrement technique directly on a multi-component signal without the filtering step, in order to provide what we called a multimode signature. The idea is then to propose an adapted method to separate the different modes in this multi-mode signature. For that purpose the modal parameters are first roughly approximated in a non-optimal way and with a bias from a time-frequency distribution. Based on this initialization, a final non-biased estimation is achieved via maximum likelihood estimator. The results conducted in this paper, over simulated and real world ambient vibrations on high rise buildings, suggest positively that the proposed method is totally autonomous and is capable of yielding very good estimation results.