The purpose of this study was attempting to predict the surface settlements and wall deflection during excavation through the feedback analysis. The optimization process using pattern search back-analyzes soil parameters. The results obtained from the back analysis were shown that the difference between predicted and measured wall deflection for next excavation stages was within 10% in average and that a rather large difference in the prediction of surface settlements. However, the location of the maximum surface settlement and the extent of the surface settlements were close to those of field measurements.
Predicting the responses of retaining structures and settlements of ground surface during stages of construction for excavation supporting system is an important task for geotechnical engineers. The observational methods (Peck, 1969) are often used in deep excavation projects where the measured data collected provides the knowledge of ability to determine whether the excavation is feasible under various methods of construction. The semi-empirical methods (Clough and O'Rourke, 1990) provide a rational and systematic means of estimating soil movements based on the stiffness of the support system. These methods may be unable in providing rational predictions of those responses of retaining structures in deep excavation. With the rapid development of computer techniques, the numerical analyses have been used and are capable of modeling serious phases of excavation such as dewatering, soil removing process, strut installation and preloading. More importantly, structures analyses of the numerical model can provide the prediction of the behavior of retaining structures and settlement of surrounding ground. However, the accuracy of the numerical results mainly depends on the choice of an appropriate soil model and the calibration of the soil parameters to the analyses. Unfortunately, the soil parameters from laboratory tests are often not reliable due to sampling disturbance, preparation, and differences in boundary conditions.
