基于大数据和深度神经网络的桩承载力时效性预测

Pile Bearing Capacity Prediction Based on Big Data and Deep Neural Network

以湛江组黏土中长期竖向受荷桩基为研究对象,根据湛江地区黏土中桩基承载力时效性测试的大数据,选取桩基几何参数、桩周土物理力学参数以及休止龄期作为深度神经网络(DNN)的输入,提出了适用于大规模数据样本的桩基承载力时效性预测模型,并利用该DNN模型与支持向量机(SVM)方法进行了对比分析。DNN预测模型比SVM方法更优,预测结果更具规律性和精准性,能满足联合多实例、大规模样本的预测要求,具备在大规模样本数据情况下的预测能力,可为计算桩基承载力时效性提供参考。

Taking the long-term vertical load pile foundation in Zhanjiang clay as the research object,according to the large data of pile bearing capacity timeliness test in clay soil in Zhanjiang area,the pile geometry parameters,pile perimeter soil physical and mechanical parameters and resting age are selected as the input of deep neural network(DNN),the pile bearing capacity timeliness prediction model applicable to large-scale data samples is proposed,and the DNN model is used to compare and analyse with support vector machine(SVM)method.The results show that the DNN prediction model is better than the SVM method,and the prediction results are more regular and accurate,which can meet the prediction requirements of joint multiple instances and large-scale samples,and have the prediction ability with large-scale sample data,and can provide reference for calculating the pile foundation bearing capacity timeliness.