Research On TBM Performance Prediction And Risk Identification Considering Multi-source Information Uncertainty

Due to the advantages of high boring efficiency,low construction impact and cost for long-distance construction,tunnel boring machine tunneling has been more and more widely used in the construction of deep-long tunnels.However,there still exists some problems and deficiencies,such as the sensitivity to geological conditions,which can cause serious casualties and large economic losses in case of accidents.Therefore,safe construction and rapid excavation under complex geological conditions is a difficult problem to be solved urgently in TBM construction,in order to give full play to the advantages of TBM construction and achieve the ultimate goal of safe and efficient excavation,how to accurately predict the tunneling efficiency of TBM and reasonably adjust the boring parameters to improve the warning and prevention ability of accidents is a hot issue in the field of TBM construction.This paper was financially supported by the National Natural Science Foundation of China(grant number:41972270):Evaluation,prediction and optimization of TBM tunneling efficiency based on the mechanism of multi-source information feedback.It was studied deeply and systemly by taking the construction with double shield TBM of Lanzhou long water conveyance tunnel project as the engineering background,and making full use of the method of field test,lab test,mathematical statistics and intelligent optimization.Novel prediction models of TBM tunneling performance consdering the uncertainty of multi-source parameters were proposed,the possible risk of TBM tuneling process and the abnormal prediction results confirm each other,which verifies that the prediction model can quantitatively explain the characteristics of uncertainty in the construction process.The research results provide a new idea for the forecasting of TBM tuneling efficiency,estimation of construction schedule as well as risk warning of construction.The main research work and contribution of this paper are as follows:(1)Based on the statistical counting of the occurrence frequency of input parameters used in various TBM performance prediction models,effective indicators for characterizing the geological conditions of the surrounding rock,mechanical performance of the TBM and construction management factors were proposed:given that the ambiguity of geological parameters,rock mass rating RMR,Cerchar abrasiveness index CAI and rock hardness H were used to measure geological conditions;for the randomness of mechanical parameters during the construction process,cutter thrust TF,cutter torque CT and rotational speed RPM were employed in the new models;and other related downtimes was proposed to quantify the uncertainty of human factors.For this purpose,the data was obtained from Lanzhou water conveyance tunnel project in China,a TBM performance database including multi-source information was established.(2)Consdering the adaptive characteristics of the beetle antennae search optimization algorithm can fit the randomness during TBM construction,and the global optimal iteration can be realized by boosted regression tree,a TBM utilization prediction model coupled with BAS method and BRT algorithm was developed,and the prediction accuracy and generalization performance of the model were reliably validated by comparison with other models.The results show that the model has good parallel processing effect and robustness,moreover,the risk indicating ability of the model was verified by comparing the feedback information of the prediction results with the typical geological risks in the tunneling process.(3)In order to quantify the uncertainty of various construction activities,the hyperparameters were adjusted adaptively by assigning different weights to these input factors,and the synergistic relationship among different input factors was improved,thus a weighted random forest model was developed to predict the TBM advance rate.The prediction results show that the model possesses high precision,strong robustness and generalization ability,and it has the advantages of fast training process,less overfitting with simple tuning and optimization steps.In addition,the possible risk of TBM tuneling process and the abnormal prediction results confirm each other,indicating that the model is a reliable method to ensure fast and safe construction of TBMs.(4)A TBM advance rate prediction model based on Bootstrap method and SVR-MKELM algorithm was proposed by introducing the idea of interval prediction.Taking the construction with double shield TBM of Lanzhou long water conveyance tunnel project as the engineering background,the rationality of input parameters was pointed out.And the validity of the developed TBM advance rate interval prediction model was verified as well.The results show that the developed interval prediction model of TBM advance rate provides a relatively accurate point prediction result,and constructs a clear and reliable AR prediction interval to cover the actual TBM advance rate completely.With the improvement of the confidence level,the uncertainty that can be contained in the prediction interval is also increasing.Moreover,the possible risk of TBM tuneling process and the abnormal interval width confirm each other,which verifies that the interval prediction model can quantitatively explain the characteristics of uncertainty in the construction process.