Prediction of TBM penetration rate based on Monte Carlo-BP neural network

Based on the BP neural network model of machine learning method, the corresponding random input parameters are generated by Monte Carlo method, and the prediction of TBM driving speed is studied. In this study, the machine learning method is applied to the prediction of TBM penetration rate, and the established empirical model has higher accuracy and practicability. After selecting the predictive control type of BP neural network, according to the control requirements of TBM, system composition and the characteristics of different geological tunneling, the appropriate data are selected to train the neural network, and the predictive control model of neural network for TBM with high convergence and real-time performance is established. Monte Carlo method has strong optimization and control functions in the realistic planning of many complex problems. In the process of TBM velocity prediction, the random input of parameters is realized by Monte Carlo method, which makes the prediction more accurate. BP neural network is used to predict the penetration rate of TBM. Its accuracy mainly depends on the accuracy of input parameters. The actual measured and predicted values of TBM driving speed are basically near the straight line x = y as the horizontal and vertical coordinates, and the correlation coefficient R = 0.9789. Therefore, the BP neural network combined with genetic algorithm has a high reference value for the prediction of TBM driving speed. When the TBM type is the same and the system equipment is the same, four factors, namely uniaxial compressive strength, Brazilian tensile strength, peak slope index, and distance between planes of weakness, are taken as input parameters of BP network by calculating the weight of influencing factors. In the specific operation, the genetic algorithm is used to iterate continuously to find the optimal solution of the initial weight parameters of BP neural network. In this study, this prediction method is applied to practical prediction. The feasibility of this method is verified by comparing with the final actual measurement result, which is of great practical significance to the evaluation of engineering, design scheme and cost control.