TH97875A - Forecast of subsurface subsidence from double tunneling with the equilibrium-balanced borehole by the neural network. - Google Patents

Abstract

------ 10/01/2018 ------ (OCR) page 1 of number 1 page invention summary. Prediction of subsurface subsidence from double tunneling with a balanced earth pressure stabilized drill head By the artificial neural network according to this invention Developed to be suitable for twin tunnel construction, it is possible to establish a relationship between the two tunnels that affect the complex subsidence. To forecast soil subsidence value ------------ Forecast of subsurface subsidence from double tunnel drilling with pressure adjustable drill head. Soil balance by a neural network according to this design. Has been developed to have Suitable for construction of double tunnels It is able to relate the factors affecting the subsidence. The complexity of the two tunnels can be used to predict the subsidence of subsidence.

Claims (2)

------10/01/2561------(OCR) Page 1 of 2 pages. Claims. 1. Prediction of ground subsidence from twin tunnel boring with balanced soil pressure tunnel boring machine using an artificial neural network. It will consist of 21 sub-units in the data layer, consisting of: (I) tunnel boring depth (kPa), (2) distance from the tunnel boring starting point from the station (T), (3) distance between tunnels (T), (4) soil properties at the tunnel top 1, (5) soil properties at the tunnel top 2, (6) soil properties at the tunnel top 3, (7) soil properties at the tunnel bottom 1, (8) soil properties at the tunnel bottom 2, (9) groundwater height from the bottom of the tunnel (T), (10) drilling head pressure (kPa), (II) drilling speed (T/T), (12) drilling head depression angle (degree), (13) grout injection pressure to fill gaps behind the tunnel walls (bar), (14) grout percentage (%), (15) drilling head pressure (kPa), (16) drilling speed (T/T), (17) drilling head depression angle (degree). (18) Pressure of grout injection to fill gaps behind the tunnel walls (bar) (19) Percentage of grout filling (%) (20) Type of drilling head (21) Tunneling time (r) with factors different from request number 0801002990, namely factor (3), tunneling factor, the second factor (15M19) and factor (21) Artificial neural network for predicting surface settlement, page 2 of 2 pages from tunneling, with a balanced soil pressure drilling head, taking into account the construction of parallel tunnels, where the number and sub-units of latent layers, as well as the calculation cycles, can be found by trial and error by dividing the selected data into 2 groups: 1) Training set, which contains approximately 25 percent of the total data, and 2) Testing set, or all data, by testing to find the network. An optimized neural network (ANN) minimizes the mean square error (RMSE) through training and testing processes. In the hidden layer, each hidden node receives factors from the input layer, which are then used in a sigmoid algorithm for calculations in the hidden layer. The results from each hidden node are then recalculated using the Sigmoid function in the output layer. This process constitutes one main iteration. The neural network then sends the results back to each hidden node for repetition, repeating the calculations for 2,000 iterations. The bias in each iteration varies, and the final results are compared to the values provided during the network's training. This is expressed as the mean square error (RMSE). Therefore, an optimal neural network architecture would consist of 21 input layers, 1 latent layer, 20 latent layers, and 2,000 iterations. 2. Forecasting of surface subsidence from twin tunnel boring with balanced soil pressure adjusting boring heads using an artificial neural network according to claim 1, which uses 21 factors in the calculation ------------ 1. The forecasting of surface subsidence from twin tunnel boring with balanced soil pressure adjusting boring heads using an artificial neural network shall consist of 21 sub-units in the receiving layer, 1 latent layer with 20 sub-units in the latent layer, and 2,000 calculation cycles. This will be considered as the network being suitable for use in forecasting surface subsidence caused by twin tunnel boring with balanced soil pressure adjusting boring heads. 2. The forecasting of surface subsidence from twin tunnel boring with balanced soil pressure adjusting boring heads using an artificial neural network according to claim 1, in the part of the 21 sub-units in the receiving layer, must consist of: (1) tunnel depth, (2) distance from the tunnel boring starting point, (3) distance between the first and second single tunnels, (4) type of soil at the tunnel top, which is (5) Type of soil at the top of the tunnel that is hard clay (6) Type of soil at the top of the tunnel that is sand (7) Type of soil at the bottom of the tunnel that is hard clay (8) Type of soil at the bottom of the tunnel that is sand (9) Height of groundwater from the bottom of the tunnel (10) Face pressure of the first single tunnel (11) Excavation speed of the first single tunnel (12) Depression angle or elevation angle of the face of the tunnel boring machine of the first single tunnel (13) Pressure of grout injection to fill the gaps behind the walls of the first single tunnel (14) Percentage of grout filling of the first tunnel (15) Face pressure of the tunnel boring machine of the second tunnel (16) Excavation speed of the second tunnel (17) Depression angle or elevation angle of the face of the tunnel boring machine of the second tunnel (18) Pressure of grout injection to fill the gaps behind the walls of the second tunnel (19) Percentage of grout filling (20) the limestone of the second tunnel, the type of tunnel boring machine, and (21) the time it took for the second tunnel boring machine to reach the position previously reached by the first tunnel boring machine.