Structural Modal Parameter Identification Under Non-white Ambient Vibration

Parameter identification is the basis of structural health monitoring. For convenient mathematical treatment, almost all methods for operational modal analysis assume unmeasured ambient excitation is stationary white stochastic process. However, ambient vibration of civil engineering structures, including heavy traffic, wind gusts and so on, behaves obviously non-stationary and non-white. As the development of signal processing technique, structural modal parameter identification under non-white ambient vibration becomes an active research topic.Time-domain stochastic state space model and stochastic subspace identification method are studied. Statistic characteristic of the white noise input is the key to establish relationship between covariance of measurements and system terms of structures, and precondition for the following algorithms. Covariance-driven and data-driven stochastic subspace identification methods have similar theoretical background of state space model but different mathematical processing techniques.In order to loose the stationary-white-noise assumption of input, introduce pre-processing filters to simulate the non-stationary and non-white ambient inputs, and establish augmented state space models, composed of pre-processing filters and original structural stochastic systems, to identify modal parameters of structure. Original structural stochastic systems are excited by non-stationary and non-white ambient inputs, and augmented systems still undergo ideal stationary white noise. Then vibration measurements have time-varying statistic characteristics because of the non-stationary input; false modes induced by the non-whiteness of input mixes in identification results; and the inherent frequency and mode shape of structures keep unaffected. Augmented state space model provides more precise theoretical sustainment for practical application of modal parameter identification methods, and foundation for researches on structural modal parameter identification under non-stationary and non-white ambient vibration.In order to eliminate effect of dominant frequencies in ambient input, the delay index is introduced to modify traditional covariance-driven and data-driven stochastic subspace identification methods respectively for stationary non-white ambient input whose covariance is not zero near the vertical axis. And the determination of delay index in practice is further studied. Then covariance-driven stochastic subspace identification method based on delay index is programmed. Numerical simulations and tests of simply supported beam bridge demonstrate that modified stochastic subspace identification method based on delay index could automatically avoid false modes induced by non-white input, looses the stationary-white-noise assumption of input, and has important significance for the parameter identification of practical civil engineering structures.