溶洞上覆土层触探特征分析

Penetration Characteristics of Karst Overburden Layer

本文首先介绍基于归一化的锥尖阻力Q_t和归一化的摩阻比F_R等触探参数的Robertson分类图及其解析表达形式,再根据研究区域的土层分布状况,简化Robertson分类图和贝叶斯模型。最后收集江西省某高速公路沿线溶洞上覆土层89个CPT和钻孔取样资料,分别采用最大似然法和贝叶斯法对Robertson分类图中的边界进行修正和比较,发现变异系数COV越小,先验分布越接近Robertson土壤分类图,预测结果越接近于Robertson土壤分类图,准确率也较以往的70%有所提高。变异系数COV越大,先验分布越为含糊不清,预测结果越接近于最大似然法的结果,但由于考虑了先验分布,准确率仍高于最大似然法。而在样本数量有限的情况下,最大似然法计算结果与Robertson分类图存在较大差别,准确率较差,应谨慎使用。

Based on the penetration parameters including normalized resistance awl produced （Qt） and friction- resistance ratio （FR）, the Robertson classification chart and analytical expression were introduced firstly. In terms of the distribution of soil layer in the study region, the Robertson classification chart and Bayesian model were simplified. Subsequently, eighty-nine CPT and information of drilling and sampling of karst overburden layer along a highway in Jiangxi province were collected. Meanwhile, the boundary of Robertson classification chart was conducted to modify and compare by using the maximum likelihood method and Bayesian model respectively. It was found that the smaller the coefficient of variation was, the prior distribution the forecast value were closer to the Robertson soil classification chart, leading to improvement of 70% of accuracy compared with the foretime. However, with the increase of coefficient of variation, the prior distribution was more ambiguous. And the accuracy of forecast value which was closer to the result of maximum likelihood method was higher than the maximum likelihood method, with the considering of the prior distribution. Under the situation of the limit sample quantity, the major difference of results between the maximum likelihood method and the Robertson classification chart was appeared. The maximum likelihood method should be used with caution due to the relatively poor accuracy.