Abstract: In order to improve the accuracy of safe operation assessment for larger aqueduct structure, a valid Hilbert -Huang Transform (HHT) modal parameter identification method based on secondary filtering under ambient excitation was proposed. For the vibration signal of large aqueduct structure, high-frequency white noise and low-frequency noise are often mixed into structure vibration information, which belongs to non-stationary and nonlinear signal in low signal to noise ratio (SNR). Wavelet threshold method can partially filter out the high-frequency white noise, and Empirical Mode Decomposition (EMD) can be used to obtain a series of intrinsic mode functions from high frequency to low frequency. So the secondary filtering combined wavelet threshold and EMD was used to process vibration signal to achieve a higher precision and a better de-nosing effect. High-frequency white noise and low-frequency water flow noise were filtered out firstly through the secondary filtering, highlighting the useful dominant dynamic characteristics of structure, which reduced the noise interference in a certain extent and avoided the phenomenon of modal frequency mixed effectively during the later modal identification. Then structure system order was determined by theory of singular entropy increment. Finally the modal parameter of large structure was identified through HHT in combination with natural excitation technique (NExT) to process the de-noised signals, which improved accuracy and precision of modal parameter identification. Taking U-shaped aqueduct of Jintai river pumping irrigation as the research object, we collected the vibration response data under normal working condition. We then used this proposed method to identify vibration response data to achieve the modal parameter of aqueduct structure. According to fluid-solid interactions theory, the three-dimensional finite element model of a cross U-shaped aqueduct structure was constructed through finite element model analysis to obtain the structure modal parameter under the same working condition. By comparing the modal parameter results from HHT model parameter identification method with three-dimensional finite element analysis, the result showed that the error between two methods was very small. The study showed that HHT modal parameter identification based on the secondary filtering method can be used to obtain the dominant structure vibration information and to improve the de-nosing precision, which decreased the strong noise disturbance and suppressed the modal frequency mixed phenomenon, and identified effectively structure modal parameter under strong noise background. Thus the method can be extended to lager aqueduct structures to improve the structure safe operation assessment. This proposed method provided a new aspect for safe operation and on-line monitoring of the aqueduct, and can be used effectively to solve the problem of structure modal parameter identification under ambient excitation, especially under the background with strong noise. Thus the method would have a broad prospect in engineering application.