砂性地层盾构掘进撕裂刀磨损规律分析

Tearing Cutter Wear Law Analysis of Shield Excavation in Sandy Stratum

目的:依托南京和燕路过江通道盾构隧道工程,针对砂性地层中撕裂刀磨损严重的问题开展研究,寻找适合砂性地层撕裂刀磨损量预测的方法。方法:实测砂性地层中盾构撕裂刀的磨损情况,分析不同换刀区间刀具磨损量;利用经验公式法对整个砂性地层撕裂刀磨损量进行拟合,利用实测磨耗量反推得到不同切削半径的撕裂刀磨耗系数;利用BP(反向传播)神经网络法对撕裂刀磨损量进行拟合;比较经验公式法和BP神经网络法适合的应用条件。结果及结论:粉细砂地层下全刀盘撕裂刀磨耗系数为1.48×10^(-3)mm/km;粉细砂-中粗砂地层下刀盘中心刀、正面刀和边缘刀区域撕裂刀的磨耗系数分别为7.84×10^(-3)mm/km、17.88×10-3 mm/km、28.64×10^(-3)mm/km;利用BP神经网络对砂性地层撕裂刀磨损规律进行拟合,效果较好,拟合优度大于0.95,误差率为20.7%。建议砂性地层中撕裂刀的磨损量预测使用经验公式法;其他地层中经验公式法预测效果较差时,可使用BP神经网络法进行刀具磨损量预测。

Objective:Based on the shield tunnel project of Nanjing Heyan Road river-crossing channel,the problem of tearing cutter severe wear in sandy stratum is studied,and the methods suitable for predicting tearing cutter wear amount in sandy stratum are explored.Method:The wear of shield tunneling tearing cutter in sandy stratum is measured on-site,and the cutter wear amount of different cutter changing intervals is analyzed.EF(empirical formula)method is applied to fit the wear amount of tearing cutters in the entire sandy stratum,and tearing cutter wear coefficients of different cutter radii are derived reversely from the field-measured wear amount.BP(backpropagation)neural network method is applied to fit the tearing cutter wear amount.The application conditions suitable for EF method and BP neural network method are compared.Result&Conclusion:The wear coefficients of full head tearing cutter in fine sand stratum is 1.48×10^(-3)mm/km;the wear coefficients of the tearing cutter in lower disc center cutter,front cutter and edge cutter areas when constructing in the fine sand-medium coarse sand stratum are 7.84×10^(-3)mm/km,17.88×10^(-3)mm/km and 28.64×10^(-3)mm/km,respectively;the BP neural network is applied to fit the wear law of tearing cutter in sandy stratum,and the results are adequate with goodness of fit greater than 0.95 and error rate of 20.7%.EF method is recommended for tearing cutter wear prediction in sandy stratum,but when the prediction performance of the EF method is poor in other strata,the BP neural network method can be employed.