PSO优化的SVM旋压触探岩土识别方法

针对现有的旋压触探测量精确度不高,没有准确的土力学模型的问题,该文提出基于支持向量分类机的岩土识别方法,并且用粒子群寻优算法来确定支持向量机的惩罚因子和核函数参数。首先,根据旋压触探勘测的实际情况,选取锥尖阻力、侧壁阻力、摩阻比、探杆倾斜角、岩土温度及湿度等参数构建旋压触探分析的指标决策体系;其次,基于粒子群寻优算法和支持向量分类机,建立岩土测试指标和土力学指标映射体系;最后在天津某地进行5次实地勘测,通过得到的静力触探和旋压触探的711组对比数据,进行该方法的验证,其分类结果准确率大于96%,符合实际工程的需要。

Rock and soil identification by using rotary penetration sounding tests based on SVM optimized by PSO

Due to the accuracy of rota penetration sounding tests is not high,and the soil mechanics model is not accurate.A rock and soil identification method based on a SVM is proposed,and using PSO to determine the penalty factor and kernel function parameters of SVM.Firstly,according to the actual situation of rotary penetration sounding tests,selecting parameters such as cone tip resistance,sidewall resistance,friction ratio,probe inclination angle,geotechnical temperature and humidity to construct index decision system for rotary penetration sounding tests.Secondly,based on the PSO and SVM,a mapping system for geotechnical testing indicators and soil mechanics indicators is established.Finally,five field surveys were carried out in a certain place in Tianjin.The verification of this method was performed by the 711 comparative data of static penetration tests and rotary penetration sounding tests.The accuracy of the classification results is over 96%,which was in line with the actual project.