Validation of a liquefaction model in calculating earthquake induced deformations of the San Fernando dams.

Abstract

The objective of this study was to examine a simple 3D liquefaction model by introducing it to the wellknown earthquake imposed San Fernando dams. The model was implemented into a finite element code called Plaxis where an uncoupled dynamic code was used. The paper seeks to highlight the model limitations by evaluating the model response and the soil response caused by the induced San Fernando earthquake motions. Three different cases of the lower San Fernando were considered, where each case increased the amount of dilative soil in the dam. The computation of the upper dam was done in attempt to see if the differing results from the lower dam are the same as the differing results from the upper dam.

The study found several model limitations which were affecting the results from the case study. One of the most important limitations of the model is that it does not have a defined void ratio dependent critical state, and thereby the state-dependent dilatancy is not considered in the model. Due to this limitation the input parameters need to be chosen extra carefully, and as further research a detailed sensitivity study of the parameters is recommended. The model was also found to have a significantly high computation time when considering a large amount of contractive soil. The results gained from thecase study were therefore not as expected. A too high amount of liquefied soil was generated which led to too high displacements. The analysis also showed that by increasing the amount of dilative soil in thedam, the displacement decreased. Therefore, one may also believe that the use of an uncoupled dynamic code was affecting the amount of excess pore pressure generated and thereby the amount ofliquefied soil. The result from the upper San Fernando dam differed in the same way as for the lowerdam. A too high amount of liquefied soil was found which resulted in too high displacement. However,further studies that either include a consolidation phase after the 20 seconds of earthquake shaking, oruses a fully coupled dynamic code are recommended. A study of this type would give a better insight inthe limitations of the model, as the dissipation of excess pore pressure and soil consolidation may have asignificant effect on the dam deformation. Although the LM3D is a simple numerical model, predicting unexpected results, it only needs some simple adjustments to get better.