| The feasibility of the use of system identification in geotechnical engineering design and construction was studied. The goal was to develop practical procedures to utilize information obtained during construction to modify an initial design to provide optimum safety and economy at any stage throughout the construction process.;A number of estimation techniques, including Bayesian estimation, maximum likelihood estimation, generalized least squares estimation, and least squares estimation, were reviewed for consistency with the amount of prior information typically available in geotechnical engineering. Different optimization algorithms were also reviewed, with particular emphasis on their efficiency with respect to number of function evaluations. Hypothetical studies were carried out to select an estimation technique and optimization algorithm for the application of system identification to practical problems in geotechnical engineering.;Four different case histories from three different areas of geotechnical engineering were used to evaluate the feasibility of system identification approach. The first case history, the construction of Oroville Dam, involved a nonlinear stress-deformation problem. Application of system identification allowed the parameters of the hyperbolic stress-strain model to be accurately identified relatively soon after instrumentation data became available. The second and third case histories, the Berthierville test embankment and the Offshore test embankment, represented the time-dependent problems. System identification allowed early identification of consolidation parameters that predicted long-term settlements very well in both cases. In the Berthierville case, the parameters agreed well with laboratory-measured parameters and with heuristic experience. In the Offshore case, some of identified parameters did not appear reasonable; possible reasons were explored. The fourth case history involved the identification of model parameters from nondestructive testing.;The application to case histories gave good results providing strong evidence that system identification approach is a valid and useful tool for geotechnical design. |
