Analysis And Prediction Of Foundation Settlement Of High-rise Buildings Under Complex Geological Conditions

The geological conditions of the construction area of high-rise buildings are becoming more and more complex,such as the constructions built in karst areas,mined out affected area and so on.Make it easy to high-rise building foundation uneven settlement,serious when will endanger the safety of the buildings,if can predict foundation settlement in advance,can avoid the accident.A high-rise building in a square including karst area and mined out affected area in Tangshan is taken as an example,analysis the karst area and mined out affected area high-rise building foundation settlement law,application of BP(Back Propagation)neural network prediction model to predict high-rise building foundation settlement,come to the conclusion as follows:(1)Foundation after dealing with the high pressure grouting in A tower in karst area and in E4 high-rise residential building in mined out affected area,in the form of pile raft foundation,the foundation settlement in front of the cap basic load increase linearly with the increase of cap settlement increase slowly and gradually stabilized,satisfies the requirement of building deformation.(2)A and B tower is in the same area,A tower is influenced by karst development and B tower is with good geological conditions,When the upper load is 450 kPa,the maximum cumulative settlement of A tower is 15.55 mm and the maximum cumulative settlement of B tower is 14.5mm.E3 and E4 high-rise residential building is in the same area,E4 high-rise residential building is influenced by mined out area and E3 high-rise residential building is with good geological conditions,When the upper load is 600 kPa,the maximum cumulative settlement of E4 high-rise residential building is 24.8mm and the maximum cumulative settlement of E3 high-rise residential building is 22.5mm.(3)By using BP neural network prediction model for A tower which is influenced by karst development and E4 high-rise residential building which is influenced by mined out area settlement prediction,in karst area prediction relative error of 0.91%-2.08%;in mined out affected area prediction relative error of 0.95% to 2.11%.(4)Through to the number of hidden layer neurons not at the same time prediction model to predict the results of analysis,obtained the suitable model of the optimal number of hidden layer neurons.In karst area when the hidden layer neurons is 3 model prediction accuracy is highest,the average relative error was 1.40%,in mined out affected area when the hidden layer neurons is 4 model prediction accuracy is highest,the average relative error was 1.41%.