为实现地铁基坑变形影响因素的客观评价,以30个基坑实例为工程背景,采用数量化理论Ⅲ分析不同因素对基坑变形的影响程度。利用不同条件下的样品得分来判断各影响因素间的耦合强度;通过构建不同输入层条件的BP神经网络来评价数量化理论Ⅲ对基坑变形影响因素筛选的准确性。分析结果表明:基坑变形的主导因素包括渗透系数、基坑深度、支撑间距和嵌固深度,重要因素包括内摩擦角、黏聚力、基坑长度、基坑宽度和地下水位,一般因素包括天然重度及支护结构刚度;基坑变形影响因素间存在一定的耦合度,且多以中、低耦合强度为主;优化BP神经网络较传统BP神经网络具有更高的预测精度。验证了数量化理论Ⅲ对基坑变形影响因素筛选的准确性,证明了数量化理论Ⅲ在基坑变形影响因素分析中的适用性和有效性,为基坑变形控制提供一定的参考依据。

In the purpose of assessing objectively the influencing factors of subway foundation pit’s deformation and improving the targeted controlling of pit deformation, we examined the influence degree of various factors using quantification theory III with 30 pit examples as engineering background. We identified the coupling strength among the factors according to sample scores under different conditions and evaluated the accuracy of selecting factors by constructing BP neural network with different input layer conditions. Results revealed leading factors inclusive of permeability coefficient, foundation pit depth, support spacing and embedded depth; important factors, namely, internal friction angle, cohesive force, length of foundation pit, width of foundation pit and groundwater level; and general factors including natural unit weight and stiffness of support structure. The coupling degrees among the influencing factors are mainly at medium and low level. The optimized BP neural network has a higher prediction accuracy than traditional BP neural network, which verifies the accuracy of the quantitative theory III in screening the influence factors of foundation pit deformation. Through this study, the applicability and effectiveness of the quantitative theory III in the analysis of the influence factors of foundation pit deformation are proved, which provides some reference for the deformation control of foundation pit.