人工冻土单轴抗压强度GA-SVM预测模型

Forecasting Model for Uniaxial Strength of Artificial Frozen Soil Based on Genetic Algorithm Support Vector Machines Method

为了预计冻结法凿井中井壁结构设计中的人工冻土单轴抗压强度,利用支持向量机在处理小样本分类学习的独到优越性及遗传算法全局并行搜索优化的特点,结合影响人工冻土单轴抗压强度因素,提出了人工冻土单轴抗压强度不同核函数的遗传支持向量机计算模型,并运用该模型预计了两淮地区第四系人工冻土单轴抗压强度。结果表明,多项式核函数的遗传支持向量机模型较高斯径向基核函数及Sigmoid核函数的遗传支持向量机模型较准确地预计人工冻土单轴抗压强度。该模型为人工冻土单轴抗压强度的预计提供了一条新途径。

In order to predict artificial frozen soil uniaxial strength which is an important parameter for the shaft design during freezing sinking, the genetic algorithm support vector machine model is put forward which can cal- culate artificial frozen soil uniaxial compressive strength, based on the artificial permafrost uniaxial compressive strength test and using genetic algorithm which has the global optimization ability to optimize the support vector machine parameter which can solve small sample. The calculated results show that polynomial kernel function genetic support vector machine is better to calculate the artificial frozen soil strength. The genetic algorithm support vector machine is a new method to predict the artificial frozen soil uniaxial strength.