基于SVR方法的注浆试验效果评价研究

Research on Effect Evaluation of Grouting Test Based on SVR Method

注浆工程属于隐蔽工程,注浆效果也不可能直观、直接检测和评价。注浆后结石体强度和浆液扩散半径是评价注浆效果的重要指标,而注浆效果又受到水泥浆液粘度、注浆压力和渗透系数等因素的影响,各因素间往往存在着复杂的关联性和非线性关系。利用水泥基注浆材料对三峡库区取样碎石土体进行注浆处理,应用支持向量回归机(SVR)方法从试验所得注浆结石体强度和扩散半径数据中随机选取若干试验样本进行训练,剩余数据进行预测,建立了注浆后结石体强度和浆液扩散半径的回归模型,提出了影响注浆效果的各因素敏感性指标排序,预测和分析了多因素影响下注浆效果的变化规律。研究结果表明:基于支持向量机的预测模型可以作为预测注浆效果的有效依据。

Grouting works is a hidden works,grouting effect can not be intuitively and directly detected and evaluated. Grouting concretion stone ＇s strength and slurry diffusion radius are important indexes to evaluate the grouting effect. The grouting effect is affected by the factors such as cement slurry viscosity,grouting pressure and permeability coefficient. There are often complex correlations and non-linear relationship among these factors. Cement matrix grouting material was used to carry out grouting treatment to the soil samples in the Three Gorges Reservoir Area. The support vector regression machine（SVR） was used to deal with the test data and predict the grouting concretion stone＇s strength and diffusion radius. Several test samples were randomly selected for training,and the remaining data were predicted. The regression model of the grouting concretion stone strength and grout diffusion radius was established,the sensitivity indexes of various factors which affect the effect of grouting were put forward,the variation rule of grouting effect under multiple factors was predicted and analyzed. The results show that predictive model based on support vector machine can be used as a valid basis to predict the effect of grouting.