Positive-Back Analysis Of Frozen Row-Piles In Deep Excavation Based On Monitoring Information

The tieback foundation engineering of Nancha suspension bridge of Runyang Changjiang Highway Bridge is a large pit. A new kind of technology of freezing and row-pole methods is applied in it. In the deep excavation, row-piles are imposed larger horizontal frost-heaving force, but the force value is intractable to ascertain exactly. It brings great danger to the safety of the foundation. Because soil's mechanical parameter and frost-heaving force have great influence on the horizontal displacement of row poles and internal force of braces, it is very important to discern them. And back analysis of displacement has offered the effective way for the discerning of these parameters.The positive-back analysis of the internal forces and deformation of the deep excavation is studied based on the finite element and BP neural network method. Training sample of BP neural network of row-pole displacement and unknown parameters is established, based on fore-analysis results. The network input is row-pole displacement, and the network output is unknown parameters. After the training of BP neural network is successful, these unknown parameters are received through the network simulation of actual measured displacement of row pole. The relation of displacement and these parameters is highly non-linear. So it has the problem of generalization of neural network for functional approximation. Uniform design method is used to get training sample in order to overcome this problem and to reduce sample count. Afterwards, the higher-sensitivity parameters are corrected to get better results, according to the sensitive of parameter for row-pole displacement. The final parameters are applied in the finite element method to forecast the safety of follow-up excavation condition.Based on the monitoring information of deep excavation, the back analysis is carried on. Then the internal forces and deformation of the exterior-protected construction are predicted and analyzed. The results show that the pit's safety is close to the limit state. Some measures must be adopted, such as reinforcement measurement, revising the design, etc.