基于SA-LS-SVM的特殊地质条件下影响地铁盾构施工效率的地层参数分析

Analysis of Formation Parameters Affecting the Shield Construction Efficiency in Complex Geological Conditions Based on SA-LS-SVM

为有效控制特殊地质条件下地铁盾构施工进度,首先对盾构施工效率影响因素进行分析,并选定关键地层参数;其次将地层参数作为输入变量,联合运用最小二乘支持向量机(LS-SVM)和模拟退火算法(SA),建立盾构施工效率预测模型,并对盾构施工效率降低引起的施工成本增加进行评估;最后采用Sobol’法对关键地层参数进行全局灵敏度分析,搜索影响盾构施工效率及施工成本的关键敏感参数,为采取预控措施提供依据。选取武汉地铁A号线某泥岩区间工程进行实证分析,结果表明:盾构施工效率预测模型的测试集平均相对绝对值误差为3.49%,机械台班费预测误差为-2.9%;各地层参数对盾构施工效率的全局敏感程度由高到低依次为:内摩擦角、泊松比、弹性模量、粘聚力、渗透系数。

In order to guarantee the shield construction progress in complex geological conditions,the factors influencing the construction efficiency were analyzed to choose the key formation arguments firstly.Then,with above selected parameters serving as the input variables,the construction efficiency prediction model based on least squares support vector machine(LS-SVM)and simulated annealing algorithm(SA)was established,and the increase in construction costs was caused by construction efficiency reduction was evaluated.Finally,the global sensitivity analysis on the key formation parameters was realized by the Sobol’method,which could be used to identify the sensitive factors.Taking a part of a mudstone interval project of Wuhan metro line A for instance,the results show that the average relative absolute value error of the test set and the prediction error of mechanical platform cost are 3.49%and-2.9%,respectively.The global sensitivity of each formation parameter to shield construction efficiency is sorted in descending order:internal friction angle,Poisson’s ratio,elastic modulus,cohesion and permeability coefficient.