摘要
为解决不停机条件难以准确获取开挖面围岩条件的问题,开发了一种基于振动时频图像的TBM隧道围岩级别感知方法,实现了对刀盘振动信息进行了全域提取,并依托西北某引水隧道通过卷积核大小、批量尺寸、迭代次数和时频变换方法等方面对围岩级别感知方法进行应用和优化,验证了方法对围岩的感知效果。研究表明,基于振动时频图像的围岩级别感知方法通过加速度传感器对TBM振动数据进行采集,并进行时频变换获得振动频谱图像,然后通过卷积神经网络对围岩级别进行预测;随着围岩条件变好振动的频率分布范围逐渐变大、频带逐渐变宽,高频部分的能量在逐渐增长,Ⅱ级围岩的频率最大值达到了400Hz左右;以LeNet网络结构为基础建立的围岩级别感知方法可准确对围岩级别进行分类,通过对模型进行优化,围岩级别感知方法的平均准确率可达96.2%。
A TBM tunnel surrounding rock classification method based on vibration time-frequency images was developed to address the issue of accurately acquiring excavation face surrounding rock conditions under non-stop conditions.This method achieves comprehensive extraction of cutterhead vibration information.Relying on an aqueduct tunnel in Northwest China,the method was applied and optimized through convolution kernel size,batch size,number of iterations,and time-frequency transformation methods,thereby verifying the method's effectiveness in perceiving surrounding rock conditions.The study indicates that the surrounding rock classification method based on vibration time-frequency images collects TBM vibration data via acceleration sensors,performs time-frequency transformation to obtain vibration spectrum images,and then predicts surrounding rock grades through a convolutional neural network.As surrounding rock conditions improve,the frequency distribution range of vibrations gradually expands,the frequency band gradually widens,and the energy in the high-frequency part gradually increases,with the maximum frequency of Grade I surrounding rock reaching approximately 400Hz.The surrounding rock classification method,established based on the LeNet network structure,accurately classifies surrounding rock grades.By optimizing the model,the average accuracy of the surrounding rock classification method can reach 96.2%.
作者
周振梁
李庆楼
郑修和
张惠兰
Zhou Zhenliang;Li Qinglou;Zheng Xiuhe;Zhang Huilan(Key Laboratory of urban uderground engineering,Ministry of education,Bejing Jiaotong University,Beijing 100044,China;ChinaRailwayFirst Survey and Design Institute Group Co.,Ltd.,Xi'an 710043,China;Xinjiang Shuifa Construction Group Co.,Lid.,Urumqi 830000,China)
出处
《土木工程学报》
CSCD
北大核心
2024年第S1期72-79,共8页
China Civil Engineering Journal
基金
中央高校基本科研业务费专项资金资助(2024JBMC046)。
关键词
TBM
刀盘振动
时频变换
图像识别
围岩感知
TBM
cutterhead vibration
time-frequency transformation
image recognition
surrounding rock perception
