Correlation Analysis of Shear Wave Velocity of Macau Soils using Nonlinear Curve Fitting and Genetic Programming

Abstract

In the past half century, a number of researchers studied the correlation between shear wave velocity and SPT-N value based on the simple power-law regression model. However, the prediction performance is not good enough to apply in practice with confidence. It may be caused by using a single measurement of SPT-N and neglecting the influence of vertical effective stress and cone resistance. Therefore, those parameters are combined with SPT-N to establish new correlation models for an existing database in Macau in this study. In this project, nonlinear curve fitting algorithm and genetic programming in Matlab software were used for the correlation analysis between the shear wave velocity with SPT-N value, vertical effective stress and cone resistance, for database obtained from the LRT project. Correlation models with different format and different soil parameters are compared in order to obtain the improved performance of model prediction. Based on the results of this study, the prediction performance is improved based on the additional soil parameters combined to form correlation models with SPT-N. Comparing between vertical effective stress and cone resistance, cone resistance has better performance when combined with SPT-N to form power law regression model than vertical effective stress. In addition, the prediction performance has successfully improved when genetic programming was applied to establish correlation of shear wave velocity. However, the complexity of the correlation models from genetic programming, the format of which cannot be assigned directly, is much more complex than the simple power law function.