摘要
分析3种典型地层中盾构推进力、推进速度、刀盘扭矩、刀盘转速4个掘进参数的变化规律.应用BP神经网络建立一个以相邻2个采样时刻的盾构推进力、推进速度、刀盘扭矩、刀盘转速共8个掘进参数为输入,地层编码为输出的地层识别模型.通过60组训练样本数据对模型进行训练,训练误差控制在8×10-7以内,并用30组预测样本数据对该模型加以预测检验,预测成功率达到93%左右.结果表明,基于盾构掘进参数的BP神经网络地层识别模型能够实现盾构掘进参数与地层之间的良好非线性映射,可以在盾构掘进施工中利用掘进参数实现对地层的在线识别.
The yariation rules of the thrust force, penetration rate, cutterhead torque and cutterhead speed were analyzed in three typical strata. By used of BP neural network the strata recognition model was established with the input of 8 parameters of shield driving from two adjacent sample time and the output of strata code. The strata recognition model was trained by 60 groups of sample data and the training error is controlled within 8 × 10^-7. Another 30 groups of sample data were also adopted for testing the accuracy of the strata recognition and the strata recognition model achieves about 93% at precision. The results show that the strata recognition model with BP neural network can realize the non linear mapping between parameters of shield driving and strata, achieving on-line strata recognition based on parameters of shield driving.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2011年第5期851-857,共7页
Journal of Zhejiang University：Engineering Science
基金
国家"863"高技术研究发展计划资助项目(2007AA041806)
国家"973"重点基础研究发展规划资助项目(2007CB714004)
关键词
盾构
推进力
推进速度
刀盘扭矩
刀盘转速
神经网络
地层识别
shield
thrust force
penetration rate
cutterhead torque
eutterhead speed
artificial network
strata recognition
