基于ICA-SVM的超长大直径钢管桩承载力预测

基于支持向量机（SVM）和独立分量分析（ICA）建立了超长大直径钢管桩极限承载力的预测模型。先采用独立分量分析FastICA算法从实际工程的超长大直径钢管桩试桩的实测数据样本中抽取相互独立的分量，这些分量不仅去除了相关性，还保持统计独立，并服从非高斯分布，能更好地表现数据间的本质结构；然后，确定支持向量机作为分类器，以抽取的独立分量作为支持向量机模型的输入参数，建立超长大直径钢管桩的承载力预测模型ICASVM_Q；最后，采用某大桥的工程数据对预测模型进行测试。结果表明，ICASVM_Q的预测效果明显优于以原始数据作为支持向量机模型输入的SVM_Q模型的预测效果。可见，采用将独立分量分析与支持向量机相结合的方法建模预测超长大直径钢管桩的承载力是可行的，ICASVM_Q模型的预测结果可用于超长大直径钢管桩承载力的设计参考，具有一定的工程应用价值。这种方法还可以用于其他领域的智能预测研究中。

Prediction for bearing capacity of large-diameter and super-long steel pipe pile based on ICA-SVM

A predictive model of ultimate bearing capacity for large-diameter and super-long steel pipe piles is established based on Support Vector Machine(SVM) and Independent Component Analysis(ICA).Firstly the FastICA algorithm of independent component analysis is imported to abstract the independent components,which can better express the essence in sample data and keep statistics independent without correlation as well as obey the Gaussian distribution,from large-diameter and super-long steel pipe pile data measured in an actual engineering.Secondly,by determining the support vector machine as classifier,whose input are the previously extracted components and whose output is the predicted bearing capacity,ICASVM_Q model is established to predict bearing capacity of large-diameter and super-long steel pipe piles.By data testing of a bridge engineering,it is shown that the predicted results of ICASVM_Q are better than SVM_Q which uses original engineering data as input of SVM model.Studies indicated that the modeling method by combining SVM and ICA used to predict ultimate bearing capacity of large-diameter and super-long steel pipe piles is feasible and the predicting bearing capacity of the ICASVM_Q model can be referred by designers in engineering.This method can also serve as a reference of intelligent forecast in other areas.