基于特征增强的盾构掘进参数显式预测分析

Explicit prediction analysis of shield tunneling parameters based on feature enhancement

盾构法施工时掘进参数的选取是保证盾构机安全、高效掘进的关键。为了实现掘进参数的预测与揭示参数之间的关联性,提出一种掘进参数的显式分析方法。依托实际的盾构隧道掘进数据,首先通过构造掘进参数的二次项以及耦合项进行特征增强;其次通过五折交叉验证和逐步线性回归相结合的方式获得最优特征子集;最终通过最优特征建立掘进参数的显式预测模型。研究结果表明:结合特征增强和特征选择可成功地将模型中的20个特征缩减到5个之内;建立的掘进参数预测模型,能良好反映真实值的变化趋势,且刀盘转速、土仓压力、总推进力、推进速度、螺机转速相对误差在15%以内,刀盘扭矩相对误差约21%,各参数误差基本可满足工况需求;此模型可定量揭示特征参数对预测量的影响程度,如对总推进力影响最大的特征是刀盘扭矩与推进速度的耦合项,其权重为1.257,而推进速度和土仓压力受到多个独立参数组成的耦合项的共同影响。

The selection of tunneling parameters is the key to ensure safe and efficient tunneling of shield machine.In order to realize the prediction and disclosure relevance of tunneling parameters,an explicit analysis method of tunneling parameters is proposed.Based on actual shield tunnel data,firstly,the quadratic term and the coupling term of tunneling parameters were constructed for feature enhancement.Secondly,the optimal feature subsets were gained by 5-fold cross-validation and stepwise linear regression.Finally,explicit predictive models of tunneling parameters were established based on best features.The results show that the feature enhancement and choose method can successfully reduce features from 20 to within 5.The predicted values of the tunneling parameters are consistent with the actual values.The relative errors of cutterhead speed,soil chamber pressure,total propulsion force,propulsion speed,screw speed are within 15%,while that of cutterhead torque is about 21%,which can basically meet the requirements of working conditions.The models quantity the influence level of feature parameters to predicted values with weight coefficients.For example,the most influential feature for the total propulsion force is the coupling term of cutterhead torque and propulsion speed,with a weight coefficient of 1.257.The propulsion speed and soil chamber pressure are jointly affected by the coupling terms composed of multiple independent parameters.