Predicting ground surface motions during an earthquake event is critical for many applications, such as structure design for seismic loading, determining displacement of landslide and evaluating instability of earth retaining wall. In engineering practice, one-dimensional equivalent linear (EL) method is widely used for ground response analysis. The EL method approximates the nonlinear soil ground using an equivalent linear system. Constant soil modulus and damping properties in the linear system are determined to be compatible with the strain levels. The method only considers the situation where horizontal shear wave vertically propagates though layered soil strata.

This thesis modifies the EL scheme and evaluates the bias and uncertainty associated with the one-dimensional as...[ Read more ]

Predicting ground surface motions during an earthquake event is critical for many applications, such as structure design for seismic loading, determining displacement of landslide and evaluating instability of earth retaining wall. In engineering practice, one-dimensional equivalent linear (EL) method is widely used for ground response analysis. The EL method approximates the nonlinear soil ground using an equivalent linear system. Constant soil modulus and damping properties in the linear system are determined to be compatible with the strain levels. The method only considers the situation where horizontal shear wave vertically propagates though layered soil strata.

This thesis modifies the EL scheme and evaluates the bias and uncertainty associated with the one-dimensional assumption. Based on the review of the EL approach, there is a need to make the soil modulus and damping properties frequency-dependent in order to enrich the soil response to high frequency components of the input motions. However, the frequency dependent equivalent linear (FDEL) method may over-amplify the ground response to high frequencies motion. Therefore, a modified FDEL method is proposed by combining the EL scheme and the FDEL scheme through a combination factor. The modified FDEL method is implemented into a well-known program SHAKE.

Several cases were conducted to benchmark the accuracy of the modified FDEL method against a fully nonlinear method. The results suggested strong motions and motions with rich portion of high frequency components would be overestimated by FDEL approach. Statistically results for different NEHRP site classes suggest that a constant combination factor 0.2 is suitable for all site classes.

With the availability of large number of downhole array data in KiK-net, validation of the one-dimensional equivalent linear methods is conducted. A total of 7440 ground response cases for 100 KiK-net stations under more than 1000 earthquake events were simulated by using the conventional EL method, the FDEL method and the modified FDEL method, respectively. Recordings from the downhole array were used as bedrock input motions in the ground response analysis, and the computed surface motions were compared with actual measurement.

Comparison of the transfer functions derived from the EL analyses with these from recorded seismographs suggested that the one-dimensional assumption is approximately valid for only 16 stations. Seismic waves may not propagate vertically due to horizontal variation of soil properties. Violation of the one-dimensional assumption results in an increase in the standard deviations of the inter- and intra-site residuals by 20% and an increase in the fixed bias by 71%.