The Prediction And Analysis Research Of The Vertical Bearing Capacity Of Cast-in-place Pile Using Neural Network

Pile foundation takes an important part in industrial and civil building.The quality of the pile foundation is directly related to the whole building safety, especially the accurate determination of bearing capacity of single pile is surely a "bottleneck" problem.Because of the uncertainty soil layer and environmental, the diversity of construction methods and the limitation of calculation models,the calculation results of the bearing capacity of pile foundation are not accurate enough that will affect the security reliability and economical rationality of the buildings.This paper probes into the current main calculation methods that used to confirm the ultimate bearing capacity of the single pile and analyzes various factors which affect the vertical ultimate bearing capacity of single pile in detail. In addition,this paper uses the method AHP to find out the main factors that influence the bearing capacity of cast-in-place pile, which provide the basis for the selection of the input nerve parameters for the neural network model.This paper introduces the basic concept of BP neural network and studies how to determinate a neural network structure, transfer function and learning algorithm for the cast-in-place pile prediction model to obtain the best neural network forecast model.Based on this, on the one hand, this paper establishes a sample library by using the collected measured data of bored pile tested. On the other hand, this paper establishes the bearing capacity prediction model respectively based on six influence parameters and based on nine influence parameters by changing numer of the input influence parameters. By means of contrast and analysis of the test values of static loading test for the pile,neural network predictive value and the calculated value using standard for pile foundation, the results show that this network prediction model has good prediction ability, namely the relative error of neural network predictive tests of piles and the test value of static loading test for piles less than the relative error of calculation value using standard and the test value. At the same time, It has been knowed that the increase of the number of input parameters of the model will has certain improvement for the prediction effect, and different distribution of soil in different regions also has certain influence on the accuracy of the prediction model.Using the established forecasting model, the paper also studies the relationship between the bearing capacity and single variation influence factor.By using matlab platform, the paper compiles a neural network forecast program for the bearing capacity of pile, developes graphic user interface. For those unfamiliar with matlab software, people can use it conveniently and directly to predict and analysis, no longer need to establishe parameters unit in the neural network program.