Abstract: It is of great significance to obtain appropriate soil parameters for efficient engineering design and safe construction in soft soils. However, different degrees of disturbance are inevitably induced in the soft soil samples in the sampling and test process due to the soft soil properties so that the accuracy of soil properties is affected. The back analysis of soil parameters from a large amount of construction monitoring data can effectively optimize and compensate the soil parameters obtained from the laboratory and in-situ testing. This paper presents the back-analysis results of soft soil parameters from the construction monitoring of a cut and cover tunnel in Taihu Lake by using the BP neurol network analysis method. The results indicates that, based on the back analysis obtained soil parameters, the deformation of the supporting structure could be accurately simulated and consistent with the monitored data.

Key words: Deep Excavations, Construction Instrumentation, BP Neural Network, Back Analysis Method