| Static pressure pipe pile technology have got rapid popularization and application in China because of its advantages such as no pollution, no vibration, low cost, high bearing capacity and so on. The vertical ultimate bearing capacity of single pile is main parameter in design of pile foundation, which could be determined by static load test, dynamic measurement method, empirical method etc. The static load test is the most reliable method to acquire the ultimate bearing capacity of single pile, but this test would be a significant amount of manpower, material and time. What is more, the number of test piles is limited. So, there is an important practical significance for engineering fields and further research to theoretically to forecast the vertical ultimate bearing capacity of single pile simply and accurately.Vertical ultimate bearing capacity of static pressure pile pipe is affected by many factors, such as pile body, soil around pile and construction conditions and so on. There exists a great complexity and nonlinearity between various influencing factors and vertical ultimate bearing capacity. Application of data mining technology on the pile foundation engineering field has launched for a long time. In recent years support vector machine (SVM) technology has become another research hot spot after artificial neural net in the field of artificial intelligence. The paper mainly did the following jobs. Firstly, collect the data from survey in the test pile sites, construction records, and data from static load tests in typical geological conditions of Liao-shen area. Then, combine SVM with grey theory, importance of variable in projection analysis, principal component analysis and other data mining technology to obtain the nonlinear mapping relationship between the physical and mechanical indexes of soil layer that surrounds the static pressure pile pipe, geometry size parameters of pile body, construction control parameters and the vertical ultimate bearing capacity of single pile, and accordingly build the predict model of vertical ultimate bearing capacity of static pressure pile pipe.(1) Combines support vector machine and grey forecasting model together to predict the vertical ultimate bearing capacity of static pressure pile pipe. The model adopts the advantage of "accumulated generating", which is in grey forecasting method, to deal with the sample and reduce the influence of random disturbance factors in original sequence. It is more regular and also avoids theoretical defect, which is existed in the grey forecasting method. Study of the paper showed that prediction accuracy of GM-SVM model was highly improved compared with single SVM model.(2) A SVM model based on the influencing factors importance evaluation was established to predict the vertical ultimate bearing capacity of single pile. This method consider grey relational analysis and variable projection importance analysis as the attribute preprocessor facilities, which could determine the vertical bearing character in typical geological conditions through the analysis results of these two methods. Then, the SVM model was established with the main factors influencing ultimate bearing capacity as input. The model effectively decreases the interference of non-essential factors, reduces the correlation between variables, and it improves model generalization ability and the predicting precision.(3) From the principal component analysis (PCA) we can see that there exists multi-correlation between the influence factors of ultimate capacity of single pile. Through PCA it concluded support vector machine input parameters, effectively screened system information, and improves rate and performance of the SVM model, the prediction model of vertical ultimate bearing capacity of static pressure pile pipe have a high accuracy. |
