Research On Stratum Identification Method Based On Shield Machine Excavation Parameters

During the tunneling process,shield machines face many difficulties,the first of which is the inaccuracy of geological information before tunneling.If geological conditions change and the shield machine’s tunneling parameters are not properly adjusted,it will lead to damage to the shield machine’s cutting tools,and even endanger the safety of the equipment.Therefore,how to accurately judge the formation state in front of the cutter head and timely avoid risks during tunneling is a difficult problem that needs to be solved urgently.The working environment of shield tunneling machines is complex,and relying solely on existing geological survey data and traditional survey methods results in a certain degree of dispersion and one-sidedness of stratum information.Once encountering adverse strata such as boulders,upper soft and lower hard strata,it is a great challenge to ensure the safety and duration of tunnel construction.Therefore,it is of important engineering practical significance to be able to use excavation parameters to identify stratum characteristics and scientifically guide the operation of shield tunneling machines.Aiming at the problem that traditional surrounding rock classification methods only consider geological information and cannot represent the difficulty of tunneling by shield machines during the excavation process,a mixed numbering expression method is used to establish a tunnel cross-section stratum classification model based on the difficulty of tunneling by shield machines.Stratums are classified and numbered according to the"Excavation effectiveness level-Excavation load stability level-Comprehensive excavation difficulty",forming a corresponding relationship between stratum numbers and stratum information.Based on this,by analyzing the correlation between the tunneling feature parameters and the stratum number,the tunneling parameters with high correlation with the stratum(cutter head torque T,tunneling speed V,tunneling specific energy SE,and tunneling capability C)are extracted.According to the classification idea,a stratum recognition model based on the KNN(K Nearest Neighbors)algorithm is established.The sensitivity was used to analyze the tunneling parameters,and the most sensitive tunneling characteristic parameter matrices[T,SE]and[T,SE,C]for the formation were obtained.They were used as inputs to the formation recognition model,with a recognition rate of 86.3%and 87.1%for the formation,respectively.Considering the low recognition rate of complex strata in the geological identification model based on the KNN algorithm,which failed to fully identify some strata with fewer geological number data samples during the actual excavation process,an objective weighting method was used to establish a geological identification model based on the entropy weight method K-nearest neighbor(EWM-KNN)algorithm.The data identified incorrectly by the KNN model was reidentified,and in terms of overall strata identification,The recognition rates of the new model for strata have increased to 95.7%and 96.4%,respectively.In terms of small sample strata recognition,the recognition rate has also increased from about 30%recognized by the KNN model to about 90%.In order to quickly respond to the dynamic data of shield tunneling machine real-time excavation based on the EWM-KNN algorithm and accurately identify it in real time,it is proposed to optimize the EWM-KNN model in four aspects:dataset search structure,nearest neighbor K-value selection,distance weight,and model learning.An improved layer recognition model based on the EWM-KNN algorithm is established.Using the KD Dimension Tree(KD Tree)mining dataset search optimization method,the data search time is reduced from N~2 to N(D-1)/D,improving retrieval efficiency.Using hierarchical cross validation to select the optimal K value for real-time updates of the dataset.The TOPSIS method is used to calculate the relative closeness between each sample and the optimal value of excavation parameters,evaluate the quality level of each parameter sample distance,and reassign weights.The semi supervised learning algorithm is used to assign"false tags"to the unlabeled tunneling data,which are reinput into the EWM-KNN classifier for stratum recognition.The recognition rate of strata has increased from 96.4%to98.6%.To visualize the current geological identification of excavation,the EWM-KNN model was successfully applied to tunnel construction.This provides an effective method for obtaining geological information using shield tunneling parameters.Provide scientific basis for adjusting excavation parameters of shield tunneling machines according to local conditions.