EnglishViews: 3 Author: Site Editor Publish Time: 2024-09-26 Origin: Site
For rockfill dams:
• The dam material and the underlying soil together form a velocity structure with high speed in the shallow layer and low speed in the deep layer ("low-speed half-space").
• The fundamental mode dispersion inversion is more inclined to analyze the velocity structure of the deep foundation, resulting in insufficient analysis accuracy for the shallow dam material.
Solution:
• Use multi-mode dispersion inversion.
Higher-order surface waves are more sensitive to changes in shear wave velocity and layer thickness in the model than fundamental surface waves (Socco & Strobbia, 2004)
Research methods:
Multimodal forward modeling of Rayleigh surface wave dispersion
Forward calculation of dispersion curves:
• Thin layer method (Lysmer, 1970; Kausel and Roësset, 1981)
• Generalized reflection-transmission coefficient method (Chen, 1993).
Discretization of underground media using horizontal thin layers
• Top is free half space
• Bottom is infinite half space
Numerically solve the constitutive equations of the horizontal layered model to obtain multimodal dispersion curves.
Verification of the correctness of the forward calculation:
A horizontal layered model is used, which is taken from Yang et al., 2024. The Rayleigh surface wave dispersion characteristics of this model are calculated by two different methods:
I. Use the finite difference time domain (FDTD) method to simulate and calculate seismic wave data, and then use the phase shift-superposition method (Park et al., 1998) to calculate the dispersion spectrum
Calculation using dispersion curve forward modeling
Verification results: The results of the two methods are highly similar. The forward dispersion curve fits the dispersion spectrum well.
The inversion process uses the Markov Chain-Monte Carlo (MCMC) inversion method.
During the inversion process, the following parameters are randomly disturbed:
• Shear wave velocity (Vs)
• P-wave velocity (Vp)
• Layer thickness
After the disturbance, the dispersion curve is generated by forward modeling. From a large number of forward modeling results, the forward model with the smallest difference from the observed dispersion curve is selected, and the corresponding model is the optimized model.
By iteratively repeating the above process, the optimized model finally obtained can approximate the real medium.
Test and analysis:
Rayleigh surface wave acquisition test
In January 2024, Chongqing Yunyang built a pumped storage power station
The lower reservoir drainage material rolling test site
The vibrating roller weighs 26 tons and travels at a speed of 2.0 m/s
Three groups of rolling tests were conducted, namely 6 dynamic pressures, 8 dynamic pressures and 10 dynamic pressures. Seismic surface wave data were collected in the intervals between each group of rolling tests.
The test site is located on the leveled hillside soil. The long side of the test site is 25 meters long, the short side is 12 meters long, and the total drainage material pile height is 1.8 meters.
The rolling material is the dam drainage material, the main component of which is the local limestone produced by blasting and excavation during the construction process, mixed with a little mudstone and soil gravel.
The rockfill and hillside soil together constitute a velocity structure with high shallow parts and low deep parts.
18 node seismometers.
The sampling rate is 1kHz.
The artificial hammer position is at the beginning and end of the survey line, with a minimum offset of 1.0 meters.
Two types of seismograph arrays were used.
1. Seismograph spacing was 0.5 m, and the total length of the survey line was 8.5 m.
2. Seismograph spacing was 1.0 m, and the total length of the survey line was 17 m.
After each rolling, multiple sets of data were collected using different positions, array directions, and seismograph spacing arrangements to reduce random interference.
The phase shift-superposition method (Park et al., 1998) was used to calculate the dispersion spectrum.
The dispersion curve was manually extracted.
Verification of inversion results
