Estimation of Extreme Value Distribution and Probability of Minor Failures in Rockfill Dam Response to Non-Stationary Seismic Excitation

To address the problem of the small failure probability of highly reliable structures such as rockfill dams, marine platforms and super high-rise buildings, in this paper, a method to estimate the extreme value distribution of the earth and rock dam response under nonstationary random seismic excitation is presented, and the small failure probability based on the SGLD model is analysed. First, a preliminary estimation of the extreme value distribution based on the kernel density estimation is made and then utilized as the raw data. Second, the main body of the extreme value distribution is obtained by combining the least squares method and a fitting process based on the SGLD model. Finally, the tail of the extreme value distribution is acquired by fitting to a quadratic equation. By using an example of earth-rock dam engineering, a comparison between the direct kernel density estimation method and the method proposed in this paper is performed. The results showed that the proposed method is closer to the extreme value distribution and probability of exceedance curves calculated by Monte Carlo simulation, which verified its effectiveness and accuracy.