铁道科学与工程学报2025，Vol.22Issue(4)：1853-1869,17.DOI:10.19713/j.cnki.43-1423/u.T20241094
基于TBM振动监测的围岩参数感知技术研究
Research on surrounding rock parameter perception technology based on TBM vibration monitoring
摘要
Abstract
The rock-breaking vibration of TBM is closely related to the state of excavated rock mass in the tunnel face.To explore the feasibility of applying TBM vibration in the field of surrounding rock perception,this paper took a TBM tunnel project in Xinjiang as an example,collected TBM main beam vibration signals under different surrounding rock conditions,used wavelet thresholding method for signal denoising,analyzed the vibration signal characteristics of different surrounding rock classes from the perspectives of time domain,frequency domain,and time-frequency,and selected the five vibration characteristic indicators with the highest correlation with rock uniaxial saturated compressive strength(UCS)and rock volume joint number(Jv)to establish a surrounding rock parameter perception dataset.Based on this,BP neural network model and SVM model were developed,respectively.Finally,the perception effect of the model was verified and compared through the validation set and test set data.The research results show that the correlation between the peak and average vibration values and the surrounding rock category has been improved by 7.56%and 12.20%after using wavelet threshold denoising method for denoising,respectively.Overall,the energy of TBM rock breaking vibration is mainly concentrated in the low-frequency range.As the surrounding rock class changes from Class IV to Class II,the vibration intensity and amplitude gradually increase,and the proportion of high-frequency component energy gradually increases.Among the twenty-six vibration characteristic parameters,the average value,peak to peak value,kurtosis,centroid frequency,and root mean square frequency have the highest correlations with surrounding rock parameters,with Pearson correlation coefficients greater than 0.7.Model testing found that the BP neural network model had prediction errors of less than 2%and 8%for UCS and Jv,respectively,which were 50%and 53%lower than the SVM model compared to the same period last year,indicating better prediction performance.The research results can provide reference and guidance for the innovation of real-time perception technology of surrounding rock parameters in TBM construction.关键词
TBM/振动监测/信号降噪/机器学习/围岩感知Key words
TBM/vibration monitoring/signal noise reduction/machine learning/surrounding rock perception分类
交通工程引用本文复制引用
李宗林,谭忠盛,周振梁,李林峰,杨旸,郑修和..基于TBM振动监测的围岩参数感知技术研究[J].铁道科学与工程学报,2025,22(4):1853-1869,17.基金项目
中央高校基本科研业务费专项资金资助项目(2024JBMC046) （2024JBMC046）
