Revised hyperbolic method for predicting settlement of soft clay ground based on weighted nonlinear regression

The hyperbolic method is a method of predicting the future settlement from the measured settlement based on the assumption that the speed of settlement decreases in the form of a hyperbola over time. However, this method forces artificial linearity during linear regression analysis because both the independent variable (time) and dependent variables (time/settlement) contain time. Thus, to overcome this problem, this study proposed the nonlinear and weighted nonlinear regression hyperbolic methods to supplement the statistical incompleteness of the original hyperbolic method based on linear regression. A hyperbolic method that applied nonlinear regression analysis and a weighted nonlinear regression hyperbolic method that applied high weight linearly over time were proposed as alternatives to the existing linear regression hyperbolic method. The accuracies of the original, nonlinear, and weighted nonlinear regression hyperbolic methods were analyzed and compared based on the time-settlement data measured from six sites with thick soft clay layers in Busan new port. The proposed methods yielded superior results compared to the original hyperbolic method. Further research on the forms of various weight functions can lead to the formulation of an optimal weighted nonlinear regression hyperbolic method.