NSVR硬岩隧道掘进机刀盘扭矩预测分析

为了避免硬岩隧道掘进机（TBM）刀盘受困，提出TBM刀盘扭矩的非线性支持向量回归（NSVR）预测模型来指导TBM掘进施工.结合吉林引松供水工程现场掘进大量数据，研究TBM刀盘扭矩与掘进参数间的相关关系，得到刀盘扭矩与围岩类别、刀盘转速和推进速度具有明显相关关系：随着围岩强度由强到弱，推进速度对刀盘扭矩的影响逐渐变弱，刀盘转速对刀盘扭矩的影响逐渐变强.基于这种相关关系，建立刀盘扭矩NSVR预测模型，并将该模型应用于吉林引松隧道工程，对按1：1划分的19 854个训练样本和19 854个测试样本的刀盘扭矩进行预测.预测结果表明：训练样本集和测试样本集的平均相对预测误差分别为11.3%和12.9%，测试样本集中相对预测误差高于60%的有516个测试样本，占测试样本集总数的2.6%.各项数据表明，在给定刀盘转速、推进速度和围岩类别条件下，建立的刀盘扭矩NSVR预测模型具有较高的预测精度.

NSVR based predictive analysis of cutterhead torque for hard rock TBM

A nonlinear support vector regression (NSVR) model was proposed for the TBM's operation to avoid the hard rock tunnel boring machine (TBM) cutterhead's jam and. Based on abundant data collected from the construction site-Water Supply Project from Songhua River in Jilin province, relevant studies about the relationship between cutterhead torque and tunneling parameters were carried out. The results show obvious correlations:the softer the rock is, the less influence of advance speed on cutterhead torque will have, and cutterhead torque is more likely to be influenced by cutterhead rotate speed. According to the results, a NSVR model was established to calculate cutterhead torque. This model was then applied in the field to calculate cutterhead torque of 19 854 training samples and 19 854 test samples. The prediction results show that, the average relative prediction error of training sample sets and test sample sets were 11.3% and 12.9%. Among all the tested sample sets, there are 516 ones whose relative prediction error are higher than 60%, accounting for 2.6%. All the data demonstrates that the cutterhead torque NSVR model can afford a relatively higher accuracy with specified cutterhead rotate speed, advance speed and rock types.