Improving Dam Deformation Analysis Using Least-Squares Variance Component Estimation and Tikhonov Regularization

Abstract

Several least-squares adjustment techniques were tested for dam deformation analysis. Deformation monitoring had only been studied horizontally. Simulated observations were first performed to compare the three methods of least-squares: observations with their initial weight, observations with the estimated variance component, and the regularization method. The best scenario was adopted for the real network. The results demonstrated that a more accurate outcome is possible by least-square variance component estimation (LS-VCE). Comparison of the estimated and simulated displacements in the LS-VCE method indicated that the difference for the two east–west and north–south components is 0.05 and 0.08 mm, while this difference in the other two modes is 1.02 and 0.67 mm for observations with their initial weight, and 0.64 and 0.16 mm for the Tikhonov regularization method, respectively. In the best method, an improvement was found in the standard deviation of the east–west, south–north components, the estimation of the major axis of the error ellipse, and the trace of the variance–covariance matrix of unknowns, in comparison with the other methods.