黄土场地静力触探参数预测分析  

Prediction and analysis of CPT parameters in loess site

作　　者：刘建维[1] 井彦林[1] 张志权[1] 林杜军 

机构地区：[1]长安大学建筑工程学院,陕西西安710061 [2]中煤西安设计工程有限责任公司,陕西西安710054

出　　处：《陕西煤炭》2017年第1期6-10,共5页Shaanxi Coal

基　　金：国家自然科学基金资助项目(No.41472267)

摘　　要：以西安某黄土场地为例,论述了静力触探测试方法的适用条件。当饱和黄土层较厚、而其下部土层较硬时,测试过程中探杆易侧向弯曲而使测试难以达到理想的深度,据此提出一种采用BP神经网络算法、依据室内试验的物理力学指标预测静力触探侧壁摩阻力及锥尖阻力的方法。通过西安市区及附近建筑场地的实测对比数据作为训练样本,建立了静力触探参数即侧壁摩阻力及锥尖阻力预测的BP神经网络模型,用实际工程对模型进行了测试。研究结果表明,室内土工试验的物理力学指标如液限、压缩模量等可作为静力触探参数的预测变量;所建立的预测模型具有实用性,这一方法为静力触探参数的确定提供了一条新途径。Taking the loess site in Xi'an as an example,the applicable conditions of CPT test method were discussed. When the saturated loess layer is thicker and the lower soil layer is hard,the testing is difficult to achieve the desired depth.Therefore the method of predicting the sidewall friction resistance and cone tip resistance based on BP neural network and physical mechanics test was proposed. Taking the measured data as training samples,BP neural network model of CPT parameter was established,and the model was tested by practical engineering. The result shows that the physical and mechanical indexes of indoor soil engineering test can be used as predictive variables of CPT parameters,the established prediction model is practical and this method provides a new way for the determination of CPT parameters.

关 键 词：静力触探 BP神经网络 锥尖阻力 侧壁摩阻力 预测 

分 类 号：TU432[建筑科学—岩土工程]