Development Of Pressure Prediction And Intelligent Speed Control Device For Soil Bin Of Shield Tunneling Machine

Effective control of the pressure of the soil bin can reduce ground uplift,subsidence and other accidents caused by shield tunneling.The pressure of the soil bin is difficult to control directly,and the speed of the screw machine is one of the important ways to realize the pressure control of the soil bin.Based on the above engineering background,this paper analyzes the correlation between shield machine excavation factors and soil bunker pressure,and builds a deep neural network prediction model.In order to facilitate the engineering application,fuzzy rules were established to obtain the screw speed corresponding to the predicted value of the pressure of the soil bin.Finally,an intelligent speed control device aimed at controlling the speed knob of the screw machine was developed to realize the intelligent control of the pressure of the soil bin.The main research content of this paper includes the following aspects:（1）This paper proposes a method of how to select tunneling factors as model input when using neural networks.By collecting the excavation factor data of the shield machine on site,the correlation analysis is carried out with the soil bunker pressure,and the excavation factors that have a correlation with the soil bunker pressure are selected as the input of the neural network,and compared with the controllable factors that other researchers often choose to use The data were grouped and compared,and the results showed that the MAE and MSE of the excavation factor data screened out through correlation analysis as the input of the neural network decreased by 42.1%,62.5%,and R_adj² increased by 20%.The proposed method lays the foundation for the establishment of neural network forecasting model.（2）In this paper,a prediction model of silo pressure based on LSTM depth neural network is proposed.Based on the Py Torch deep learning framework,the advantages and disadvantages of the BGD,MBGD and SGD gradient descent in the establishment of Earth silo pressure prediction models were compared and analyzed,taking into account the practical engineering application aspects,to further explore the optimization degree of different optimizers on gradient descent,a LSTM depth neural network prediction model was established,and PSO was used to optimize the model.MAE and MSE were reduced by 36.4%and 53.3%compared with those before optimization,respectively,R²is up 6.9%.The proposed model lays a foundation for controlling the pressure of soil bin.（3）This paper proposes a fuzzy reasoning-based Wang-Mendel algorithm to extract and establish fuzzy rules from historical data.Compared with the actual engineering data,the error of MAE and MSE of the screw machine speed corresponding to the predicted value of the soil silo pressure calculated by this method is 0.065rpm and 1.528%,and the R² is 0.82,which shows that the effect of this method is better.The proposed method lays the foundation for establishing an intelligent control device.（4）A set of intelligent control device aiming at controlling the rotating speed knob of the screw machine is established.The speed knob of the screw machine in the control room of the shield machine is used as the control object,the Arduino single-chip microcomputer is used as the main control board,and the Simple FOC board with the FOC control algorithm as the core is used as the driver to drive the brushless DC motor,and the PID parameters in the FOC control algorithm Adjustment reduces motor lag,vibration,overshoot and other problems;builds a 3D model and uses 3D printing to make a physical device,and finally compares the operating value of the device with the control target value through an engineering case.The MAE and MSE errors of the device are 0.015rpm and 0.392%,R² is 0.95,indicating that the device can accurately reach the control target position,and the effect is good.