TBM Construction Envelope Rock Excavatability Classification And Tunneling Speed Prediction Under Small Section Soil And Rock Combination Geological Conditions

Because of its high efficiency,high safety and low cost,TBM construction has become the first choice for long-distance rock tunnel construction in water conservancy and transportation projects.This complex geological condition is encountered during the construction of TBM,which requires the adoption of appropriate tunnelling parameters according to the type of rock in the tunnel;if the tunnelling parameters do not match the geological conditions,the tunnelling efficiency will be reduced.In the process of TBM construction,this complex geological condition requires the use of appropriate tunnelling parameters according to the type of tunnel rock,if the tunnelling parameters do not match the geological conditions,the tunnelling efficiency will be reduced.Especially for small diameter tunnels,as the TBM body is small and light,on the one hand,if the boring thrust is small,the cutter torque is small and the penetration degree is small,the boring efficiency is not high;on the other hand,if the boring thrust is large,the cutter torque is large and the penetration degree is large,it will cause the TBM body axis deviation is large,which requires constant correction of deviation and the boring efficiency is low.Therefore,in this complex geological condition,the traditional surrounding rock classification method is no longer applicable,and a new surrounding rock classification suitable for TBM construction needs to be determined and the surrounding rock classification accurately identified.At the same time,with the development of intelligent engineering construction and the establishment of big data platforms,accurate prediction of TBM boring performance parameters has also become one of the key issues and an inevitable requirement for intelligent control and eventual realization of driverless TBM equipment.To address the above issues,the main research work and results of this paper are as follows:(1)Based on the actual data of TBM construction in western Anyang water transfer project,this paper establishes the criteria of rock excavability grading for TBM construction under geological conditions of soil and rock combination in small cross-section by using the comprehensive index of digging performance unit penetration thrust(FPI)and unit penetration torque(TPI);and proposes the PCA-RF model to identify the rock excavability grading and discusses it in comparison with BP,SVR and RF models.The results show that: 1.the established construction excavability classification criteria for TBM construction in small-section earth and rock combination surrounding rocks are applicable and overcome the limitations of the traditional surrounding rock classification method under earth and rock combination surrounding rocks;2.the recognition accuracy of PCA-RF recognition model for construction excavability classification of TBM construction in small-section earth and rock combination surrounding rocks reaches 98.3%,which is higher than BP,SVR and RF models and can meet the engineering construction needs.(2)A multi-step prediction model of TBM tunneling speed based on EWT-ICEE MDAN-SSA-LSTM hybrid model is proposed.First,four data sets are selected among different envelope classification criteria,and the original data are preprocessed using t he binary discriminant function and the 3σ principle,followed by decomposition of the preprocessed data using the empirical wavelet variation(EWT)to obtain several subs equences and residual sequences,and the residual sequences are decomposed again by the improved adaptive noise fully ensemble empirical modal decomposition(ICEEMD AN).Finally,several subsequences are substituted into the sparrow search algorithm(S SA)optimized long and short term memory(LSTM)network for multi-step training a nd prediction,and the prediction results of each subsequence are summed to obtain th e final result.The comparison with the existing model calculation shows that the perf ormance of the proposed prediction method outperforms other models,and the average accuracy reaches 99.06%,98.99%,99.07% and 99.03% from the first step prediction to the fifth step prediction in four data sets,indicating that the method has high multi-step prediction performance and generalization ability,which can provide reference for other projects.