Tunnel Boring Machines (TBMs) have been the main equipment for tunneling and underground construction due to their high safety performance and tunneling efficiency. However, the unknown and changing geological conditions during construction pose a challenge to TBM construction. As one of the essential parameters of rock properties, accurate acquisition of uniaxial compressive strength (UCS) is crucial for TBMs to adapt to changing ground conditions in a timely manner. Therefore, this study proposes a Catboost intelligent model based on Bayesian Optimization to predict UCS. Rock mass are velocity information and key TBM operational parameters are used as model input variables. The Gaussian data augmentation method is used to compensate for the difficulty of obtaining field data in large quantities. The Zhujiang Delta Water Resources Allocation Engineering field data are used in the model, and the obtained evaluation indicators MAPE, RMSE, VAF and a20-index are obtained as 9.91%, 499.38 MPa, 90.7% and 0.95, respectively. In addition, another project is selected to verify the applicability of the model. The validation results also confirm that the model is valid and reliable when applied to practical engineering.

Cutter life is an important economical index for tunnel boring machines (TBM) excavation, and its prediction is widely concerned. This paper introduces a method for predicting cutter life on the basis of statistics and regression. Traditional researches only evaluate the average mileage or time interval of each cutter change. Differently, the proposed method can accurately predict the mileage position of each cutter relatively, on the premise of knowing the installment radius of each cutter and the rock properties along the tunnel. In this procedure, the influence of the rock classification and rock properties of each cutter passing area on their passing distance is obtained by linear regression. For proposing and verifying the prediction method, totally 1,200 records of cutter changing caused by normal cutter wear are collected from the 4th Section of Water Supply Project from Songhua River. Among them, randomly selected 920 samples are used to determine the regression coefficients involved in the method. The method is verified by the rest 280 samples, and a reliable predicted result is obtained (MAPE = 27%, and R² = 0.69).

Rapid and accurate judgment of the rock mass condition is the key to guaranteeing the safety and efficiency of tunnel boring machine (TBM) tunneling. This paper proposes a method for evaluating rock mass parameters based on K-means clustering, grouping tunneling areas according to the values of TBM tunneling parameters. A dataset including rock mass and TBM tunneling data is treated by logistic normalization and principal component analysis (PCA), and large volumes of tunneling data with different features are transformed into appropriate volumes of dimensionless data. K-means clustering is used, samples are grouped according to the values of tunneling data, and the specific ranges as defined by clustering are regarded as the unified evaluated results of each group. Based on the C1 part of the Pearl Delta water resources allocation project, 100 training samples and 30 testing samples were field-collected, and the proposed method was realized by the training samples and verified by the testing samples. The evaluation accuracies of uniaxial compressive strength (UCS), and joint frequency (Jf) were 90%, and 86.7% respectively, demonstrating that the evaluation had acceptable values, and the proposed method was greatly helpful for judging rock conditions.

The utilisation of particle flow code to establish discrete element models represents an effective approach for addressing the issue of discontinuous media. This methodology has been employed by numerous scholars to analyse the mechanical properties and damage laws of geotechnical materials. However, the complex nature of the particle action mechanism within the discrete element model necessitates a considerably longer time frame for the completion of an elaborate simulation experiment than that required for a laboratory test. This presents a significant challenge for researchers seeking to investigate the mechanical properties of a large number of geotechnical materials through the discrete element method. In order to accelerate the prediction of mechanical properties for various specific discrete element models, a mathematical model of the geotechnical micro-parameters and the geotechnical strength macro-parameters has been developed using an orthogonal design considering interactions and a back propagation neural network based on Bayesian regularisation. The geotechnical strength macro-parameters, such as compressive strength and tensile strength, can be derived directly from the geotechnical micro-parameters of the discrete element models through this mathematical model. The results show that the trained network model demonstrates an aptitude for predicting the uniaxial compressive strength, tensile strength, cohesion, and friction angle of geotechnical materials. The mean square error is 11.611 for the training set and 14.207 for the test set. In the test set, the median deviation rates of the predicted values of the four strength macro-parameters from the target values are 3.90%, 4.82%, 4.30%, and 7.30%.