基于神经模糊推理系统的盾构施工地表沉降预测

盾构隧道施工引起的地表沉降,主要受盾构掘进参数和地层条件的影响,且各参数间关系复杂.已有地表沉降预测方法大都没有直接考虑掘进参数的影响,难以满足盾构快速施工超前预测预报和环境影响控制的需求.自适应神经模糊推理系统(ANFIS)是一种基于神经网络的模糊类智能模型,通过减法聚类数据细分技术自动生成模糊规则,使网络的节点和权值具有明确的物理意义,集成了神经网络数据自适应能力和模糊系统知识表达性能,特别适合于多元非线性系统的预测预报.结合北京地铁14号线东风北桥站至京顺路站区段工程实测数据,选取埋深、洞顶覆土标贯值、土仓压力、推进速度、刀盘转速、扭矩、盾构推力,以及同步注浆量为输入变量,建立了地表最大沉降量预测模型.计算结果表明,该模型计算量小,泛化能力强,计算精度高.研究成果为盾构施工地表沉降预测预报提供了新的技术方案.

Prediction of ground surface settlement induced by shield tunneling construction based on neural fuzzy inference system

Ground surface settlement induced by shield tunnel construction is affected by stratum conditions and shield tunneling parameters.The relationship between factors is complex.The traditional methods are hard to meet the prediction and forecasting demand for the safety monitoring demand during the process of shield construction because of the lack of direct consideration of tunneling parameters.Adaptive Neural Fuzzy Inference System (ANFIS) is a kind of fuzzy smart model based on neural network and the fuzzy rules are generated automatically via the subtractive clustering data subdivision technology so that the well-defined physics meaning can be obtained for network node and weight.In addition,ANFIS has a good ability of data adaptive and fuzzy knowledge representation,which is suitable for prediction and forecasting of multivariate nonlinear system.Taking the section tunnel between Dongfengbeiqiao Station and Jingshunlu Station of Beijing Subway Line 14 as an example,factors such as buried depth,standard penetration,earth pressure of chamber,rotating speed of cutter head,advance rate,the torque variation,shield tunneling thrust and synchronous grouting are selected as the input variables so that the prediction model for the maximum settlement of ground surface is established.The results show that the model has the characteristic of small amount of calculation and strong generalization ability and high precision computation,which provides a new technical solution for the surface settlement prediction and forecasting.