姚兆明 左维亚 高亚飞

摘 要：土体在低温下的冻胀将会对工程设施建设和运营等方面造成不利的影响，由于土体冻胀率受制于众多因素，因此土的冻胀率计算一直是研究的热点、难点。为了解决冻胀率计算难、繁等问题，以69组已知冻胀率样本为研究对象，采用支持向量机回归模型计算方法，随机选取59组样本对建立的径向基核函数和多项式核函数两种预测模型进行训练，再利用训练后的两种预测模型对10组样本进行预测。通过试验值和两种模型训练值、预测值进行对比，结果表明径向基核函数预测模型的预测效果更好。建立的冻胀率预测模型将支持向量机回归模型在人工冻土方面应用进行了拓展，为冻胀率预测提供了一种新的智能方法。

关键词：冻胀率;支持向量机回归模型;核函数

中图分类号： TU443  文献标志码：A

文章编号：1672-1098（2021）01-0013-05

收稿日期：2020-08-26

基金项目：陕西省煤田地质集团有限公司科研项目（SMDZ-2019CX-7）

作者简介：姚兆明（1975- ），男，安徽黄山人，教授，博士，研究方向：土体本构理论和岩土数值分析研究。

Intelligent  Calculation  Method  for  Frost  Heave  Rate of  Artificial Frozen Soil  Based on Small Samples

YAO Zhaoming1，ZUO Weiya1，GAO Yafei2

（1.School  of Civil Engineering， Anhui University of Science and Technology， Huainan Anhui 232001，China; 2. Shaanxi 185 Coalfield Geology Co.，Ltd.，Yulin Shaanxi 719000， China）

Abstract：Frost heaving of soil under low temperatures will impose negative impact on engineering facilities， construction and operation. As the frost heaving rate of soil is subject to many factors， its calculation has always been an important and challenging research topic for scholars. For this reason，69 groups of samples with the known frost heaving rates were selected in this paper as the research objects and the support vector machine  regression model was applied in the calculation of frost heaving rate. 59 groups of samples were used to train two prediction models， which were then used for the prediction of 10 groups of samples. The comparison between experimental value and the training value and prediction value of the two models shows that the radial basis function prediction model is able to give a better prediction result. The established frost heaving rate prediction model， as an expanded application of support vector  machine  regression model in artificial ground freezing， provides a new intelligent method for the prediction of frost heaving rate.

Key words：frost heaving ratio; support vector  machine  regression model; kernel function

凍胀是寒区工程面临的主要工程病害，因此冻土冻胀变形的研究对工程建设和维护具有深远意义[1]。国内外学者对此进行了深入研究。文献[2]考虑产生冻胀的各种因素，提出了一维条件下计算土体冻胀量的方法。文献[3]采用灰色关联理论分析了含水量、有机质、含盐量、土质和干密度对天山达板乔尔玛至那拉提公路延线冻土冻胀率的影响。文献[4]233分析了含水量、初始干重、冻透率、地下水位、可塑性指数对冻胀率的影响，建立了预测冻胀率的统计模型。……