基于PCA-RVM的静压管桩单桩极限承载力预测

Prediction of Ultimate Bearing Capacity of Statically-Pressured Pipe Pile Based on PCA-RVM

为解决静压管桩单桩极限承载力难以获取的难题,提出一种基于主成分分析(Prineipal Component Analysis,PCA)的相关向量机(Relevance Vector Machine,RVM)静压管桩单桩极限承载力预测方法。通过PCA将13个静压管桩常规影响因素降维成6个独立变量借助RVM模型建立独立变量与极限承载力之间的非线性映射关系,能够对仅知道影响因素的新样本进行精准预测。采用PCA-RVM方法进行预测并与BP神经网络模型对比,结果表明:PCA-RVM预测模型通过分析各因素的相关性与贡献率,将13个影响因素合理转化为6个独立变量,在信息筛选方面明显优于BP神经网络模型。在承载力预测结果的相对误差及均方差方面,PCA-RVM预测模型均占据极大优势。可见PCA-RVM模型具有精度高、离散性小等优点,为静压管桩单桩极限承载力预测开辟了一种新方法。

In order to solve the problem that the ultimate bearing capacity of static pressure pipe piles is difficult to obtain,a method for predicting the ultimate bearing capacity of static pressure pipe piles by relevance vector machine(RVM)based on principal component analysis(PCA)is presented.PCA was used to reduce 13 conventional influencing factors of static pressure pipe piles into 6 independent variables,and the nonlinear mapping relationship between independent variables and ultimate bearing capacity was established by RVM model,so as to accurately predict the new samples with only known influencing factors.The PCA-RVM method was used for prediction and compared with the BP neural network model.The results show that the PCA-RVM model which analysis the correlation and contribution rate of each factor can reasonably transform the 13 influencing factors into 6 independent variables.lt is obviously better than the BP neural network model in information screening.In terms of relative errors and mean square errors of bearing capacity prediction results,PCA-RVM prediction model has a great advantage.lt can be seen that the PCA-RVM model has the advantages of high precision and small discreteness,which opens up a new method for predicting the ultimate bearing capacity of static pressure pipe pile.