基于岩体聚类分级的TBM掘进参数预测方法

TBM tunneling parameters prediction method based on clustering classification of rock mass

复杂地质条件下TBM掘进参数的精准预测可及时优化调整掘进参数,有效指导设备施工。根据吉林引松供水工程TBM3标段的现场掘进数据,首先利用最小二乘法对TBM掘进参数和现场跟踪的岩体力学参数进行回归分析,实现由设备参数向岩体信息的转化;而后利用k-means方法对所估计的岩体力学参数进行聚类分级,建立不同围岩等级的岩机数据库;最后以对应围岩等级下TBM岩机数据作为模型输入、运行或者控制参数作为模型输出目标,利用基于极限学习机(ELM)的机器学习算法构建与围岩等级相匹配的TBM预测模型,其预测值可很好地拟合实测数据的变化趋势,平均误差在12%以内。结果表明,基于岩体聚类分级的TBM掘进参数预测方法可显著改善围岩等级多变条件下TBM掘进参数预测精度低、鲁棒性差的问题。

Accurate prediction of TBM tunneling parameters under complex geological conditions can optimize and adjust the tunneling parameters in time to effectively guide equipment construction.According to on-site tunneling data from the 3rd TBM section of Songhua River water supply project in Jilin province,this paper first uses least square method to perform regression analysis on TBM tunneling parameters and on-site tracked rock mechanics parameters,which realizes the transformation from mechanical parameters to rock mass information.Then k-means method is used to classify the estimated rock mechanics parameters to establish a database including rock mass properties and machine parameters under different surrounding rock grades.Finally,the TBM rock mass parameters and machine parameters corresponding to surrounding rock grades is used as the model input,operating or control parameters as the model output target,and ELM-based machine learning algorithm is utilized to construct predictive models matching the surrounding rock grade.The predictive value fits the change trend of the measured data well,and the average error is less than 12%.The results show that TBM tunneling parameters prediction method based on clustering classification of rock mass can significantly improve the problems of low prediction accuracy and poor robustness of TBM tunneling parameters under the dynamic change of rock mass.