A TBM intelligent decision-making system and method based on deep learning

Abstract

The present invention discloses a TBM intelligent decision-making system and method based on deep learning, comprising: a testing and acquisition module for acquiring seismic wave test data, induced polarization test data, advance drilling test data, tunneling loss, and tunneling parameters; a data set processing module for obtaining a surrounding rock feature vector based on the seismic wave test data, induced polarization test data, and advance drilling test data, and obtaining a tunneling loss vector and a tunneling vector based on the tunneling loss and tunneling parameters; and an intelligent decision-making module for receiving a preset tunneling loss range, tunneling parameter range, and tunneling mode, and using a trained deep learning model to obtain tunneling loss under different tunneling parameters based on the surrounding rock feature vector and the corresponding tunneling loss vector and tunneling vector, and selecting tunneling parameters based on the tunneling loss range and tunneling mode. The system considers the surrounding rock characteristics ahead of the tunnel face, increases the shield machine's perception of the surrounding rock characteristics, and reduces resource consumption and project costs while meeting project schedule.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel construction, and in particular to a TBM intelligent decision-making system and method based on deep learning. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The TBM construction method, centered around a hard rock tunnel boring machine (TBM), uses manually operated information-based equipment to excavate tunnels. Compared to traditional drill-and-blast tunneling methods, TBMs offer faster excavation speeds, better results, minimal environmental impact, and safety and environmental friendliness. Overall, they offer significant advantages, including high economic and social benefits. TBM construction offers significant advantages for deep, long tunnels exceeding ten kilometers, where simultaneous segmented, multi-head drilling and blasting methods are difficult to implement.

[0004] During TBM excavation, operating parameters must be adjusted in real time to adapt to changing geological conditions. Traditionally, these parameters rely primarily on the operator's experience. Operators continuously monitor TBM excavation data and, based on current operating parameters, adjust the next parameters. These parameters are adjusted based on the interaction between the tunnel strata and the TBM to ensure safe and efficient advancement.

[0005] However, this method relies too much on the experience of the operators and is not adaptable enough to geologically complex areas. Once encountering stratum changes or complex geological conditions, if the operators fail to make effective adjustments to the excavation parameters in a timely manner according to the stratum conditions, it may cause damage to the cutter, unstable excavation, collapse, machine jamming, and extended construction period.

[0006] At present, some researchers have used machine learning technology to judge the surrounding rock conditions during the TBM excavation ascent, predict the surrounding rock state parameters, and then determine the TBM excavation parameters. However, the TBM excavation ascent duration is short, and the TBM excavation distance during this period is short (generally less than 10 cm). This method still has the defect of insufficient response to stratum changes or complex geological conditions.

[0007] At the same time, some technologies currently use deep learning technology to build models to characterize the relationship between surrounding rock conditions and TBM excavation parameters; however, this method is based on learning from operators, and what is learned is the operator's operating experience, and it does not find the most suitable parameters for TBM excavation. Summary of the Invention

[0008] To address the above issues, the present invention proposes a TBM intelligent decision-making system and method based on deep learning, which fully considers the characteristics of the surrounding rock in front of the tunnel face, increases the shield machine's perception of the surrounding rock characteristics, enables the TBM to adapt to complex geological conditions, and reduces resource consumption and project costs while meeting the project schedule.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] In a first aspect, the present invention provides a TBM intelligent decision-making system based on deep learning, comprising:

[0011] The testing and acquisition module is used to conduct tunneling tests on the surrounding rock in front of the tunnel face and obtain seismic wave test data, induced polarization test data, advance drilling test data, tunneling losses, and tunneling parameters;

[0012] a data set processing module configured to obtain a surrounding rock feature vector based on seismic wave test data, induced polarization test data, and advance drilling test data, and to obtain an excavation loss vector and an excavation vector based on excavation loss and excavation parameters, respectively;

[0013] The intelligent decision-making module receives a preset excavation loss range, excavation parameter range and excavation mode, and is configured to use a trained deep learning model to obtain the excavation loss under different excavation parameters based on the surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, and select the excavation parameters according to the excavation loss range and excavation mode.

[0014] As an optional implementation, the testing and acquisition module includes a seismic wave testing module, which is used to perform advance detection of the surrounding rock in front of the tunnel face to obtain longitudinal wave velocity, shear wave velocity, Poisson's ratio and density, and thereby obtain a seismic wave characteristic vector.

[0015] As an optional implementation, the testing and acquisition module includes an induced polarization testing module, which is used to detect the spatial distribution characteristics of the surrounding rock resistivity and obtain the induced polarization characteristic vector.

[0016] As an optional implementation, the testing and acquisition module includes an advance drilling testing module; the advance drilling testing module drills and samples the surrounding rock in front of the face to obtain rock quality indicators, joint occurrence, joint surface density, joint surface thickness, joint surface filling and cementation characteristics and joint surface morphological characteristics, and thereby obtains a joint characteristic vector.

[0017] As an optional implementation, the advance drilling test module further includes performing a point load test and a friction test on the surrounding rock in front of the tunnel face, and obtaining a rock characteristic matrix based on the point load test results and the wear resistance test results.

[0018] As an optional embodiment, the testing and collection module includes an excavation loss collection module, which is used to collect resources consumed in maintaining cutterhead rotation, resources consumed in maintaining cutterhead thrust, resources consumed in maintaining cylinder pressure, and cutter wear.

[0019] As an optional implementation, the testing and collection module includes a tunneling parameter collection module, and the tunneling parameter collection module is used to collect tunneling speed, tunneling thrust, cutterhead rotation speed and cutterhead thrust.

[0020] As an optional embodiment, the excavation mode includes an energy-saving mode, which selects an excavation mode with minimal excavation loss while ensuring that the excavation speed meets the advancement speed requirement. In the energy-saving mode, an optimal excavation vector is selected through an optimization algorithm, and a minimum excavation speed is set to minimize the excavation loss while meeting the minimum excavation speed requirement. Specifically, the energy-saving mode includes:

[0021] (1) Generate an initial decision plan;

[0022] (2) Applying the trained deep learning model to the initial decision-making scheme to obtain the initial array of tunneling speed and tunneling loss;

[0023] (3) Filtering out the array that meets the minimum excavation speed requirement from the initial array, and then filtering out the corresponding excavation parameter array according to the excavation loss;

[0024] (4) Randomly select any three data from the tunneling parameter array and perform differential analysis repeatedly to obtain a candidate decision solution;

[0025] (5) Applying the trained deep learning model to the candidate decision-making schemes to obtain an array of tunneling speed and tunneling loss;

[0026] (6) Repeat steps (4)-(5) and select the decision plan corresponding to the minimum excavation loss from the new decision plans obtained.

[0027] As an optional implementation, the excavation mode also includes a rapid advancement mode. The rapid advancement mode is a excavation mode that selects the fastest excavation speed within an acceptable excavation loss range. In the rapid advancement mode, an optimal excavation vector is selected through an optimization algorithm; a maximum excavation loss is set to maximize the excavation speed while meeting the maximum excavation loss requirement. Specifically, the following steps are performed:

[0028] (1) Generate an initial decision plan;

[0029] (2) Applying the trained deep learning model to the initial decision-making scheme to obtain the initial array of tunneling speed and tunneling loss;

[0030] (3) Filtering an array that meets the tunneling loss requirement from the initial array, and then filtering out the corresponding tunneling parameter array according to the tunneling speed;

[0031] (4) Randomly select any three data from the tunneling parameter array and perform differential analysis repeatedly to obtain a candidate decision solution;

[0032] (5) Applying the trained deep learning model to the candidate decision-making schemes to obtain an array of tunneling speed and tunneling loss;

[0033] (6) Repeat steps (4)-(5) and select the decision plan corresponding to the maximum excavation speed from the new decision plans obtained.

[0034] In a second aspect, the present invention provides a TBM intelligent decision-making method based on deep learning, comprising:

[0035] Conduct tunneling tests on the surrounding rock in front of the tunnel face and obtain seismic wave test data, induced polarization test data, advance drilling test data, tunneling losses, and tunneling parameters;

[0036] The surrounding rock characteristic vector is obtained based on the seismic wave test data, the induced polarization test data and the advance drilling test data, and the excavation loss vector and excavation vector are obtained based on the excavation loss and excavation parameters respectively;

[0037] Receive preset excavation loss range, excavation parameter range and excavation mode;

[0038] According to the surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, the trained deep learning model is used to obtain the excavation loss under different excavation parameters. The excavation parameters are selected according to the excavation loss range and excavation mode.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention proposes a deep learning-based TBM intelligent decision-making system and method, which supports both energy-saving and rapid advancement modes, adapting to different engineering needs and reducing resource consumption and engineering costs while meeting engineering schedules.

[0041] The present invention proposes a deep learning-based TBM intelligent decision-making system and method that fully considers the characteristics of the surrounding rock in front of the tunnel face, increases the shield machine's perception of the surrounding rock characteristics, enables the TBM to adapt to complex geological conditions, and reduces mechanical losses and construction delays caused by improper response to geological conditions.

[0042] The present invention proposes a TBM intelligent decision-making system and method based on deep learning. During the application process, TBM continuously collects data and expands the database. When the amount of data increases to a certain level, the model is retrained to improve decision-making accuracy.

[0043] This invention abandons the previous method of learning from TBM experience and uses deep learning to construct the relationship between surrounding rock parameters, TBM excavation parameters and excavation losses. Under the premise of given excavation tendency, an optimization algorithm is used to find the parameters that best suit the current TBM state and surrounding rock state.

[0044] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0046] Figure 1 This is a logical block diagram of the deep learning-based TBM intelligent decision-making system provided in Example 1 of the present invention;

[0047] Figure 2 The advance drilling arrangement diagram provided in Example 1 of the present invention;

[0048] Figure 3 A schematic diagram of the values ​​of the induced polarization eigenvector provided in Example 1 of the present invention;

[0049] Figure 4 This is a flowchart of the algorithm of the intelligent decision-making module provided in Example 1 of the present invention;

[0050] Figure 5 Flowchart of the deep learning-based TBM intelligent decision-making method provided in Example 1 of the present invention;

[0051] Among them, drilling position 1, No. 1; drilling position 2, No. 2; drilling position 3, No. 3; drilling position 4, No. 4; drilling position 5, No. 5. DETAILED DESCRIPTION

[0052] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0054] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0055] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0056] Example 1

[0057] This embodiment provides a deep learning-based TBM intelligent decision-making system. By acquiring the surrounding rock characteristics ahead of the tunnel face, including rock strength, surrounding rock wear resistance, surrounding rock wave velocity, surrounding rock fragmentation, and water content, it detects changes in surrounding rock characteristics and provides TBM excavation parameters based on the surrounding rock characteristics and taking into account engineering needs (excavation speed and energy saving), thereby assisting the TBM in rapid and efficient excavation.

[0058] like Figure 1 As shown, specifically including:

[0059] The testing and acquisition module is used to conduct tunneling tests on the surrounding rock in front of the tunnel face and obtain seismic wave test data, induced polarization test data, advance drilling test data, tunneling losses, and tunneling parameters;

[0060] a data set processing module configured to obtain a surrounding rock feature vector based on seismic wave test data, induced polarization test data, and advance drilling test data, and to obtain an excavation loss vector and an excavation vector based on excavation loss and excavation parameters, respectively;

[0061] The intelligent decision-making module receives a preset excavation loss range, excavation parameter range and excavation mode, and is configured to use a trained deep learning model to obtain the excavation loss under different excavation parameters based on the surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, and select the excavation parameters according to the excavation loss range and excavation mode.

[0062] In this embodiment, the testing and acquisition module includes a TSP seismic wave testing module, an induced polarization testing module, an advance drilling testing module, a tunneling loss collection module, and a tunneling parameter collection module;

[0063] Specifically:

[0064] The TSP seismic wave testing module includes a TSP305 seismic wave detector, which is used to perform advance detection of the surrounding rock in front of the tunnel face to obtain the longitudinal wave velocity, shear wave velocity, Poisson's ratio and density changes of the surrounding rock.

[0065] The induced polarization test module includes an induced polarization instrument for detecting the spatial distribution characteristics of the surrounding rock resistivity.

[0066] The advanced drilling test module includes an RDP-150C drilling rig, a point load tester, a Cerchar friction tester, and a measuring ruler;

[0067] The RDP-150C drilling rig is used to drill and sample the surrounding rock in front.

[0068] The point load tester is used to perform a point load test;

[0069] The Cerchar friction tester is used to perform friction tests and calculate the wear resistance of the surrounding rock samples in front;

[0070] The measuring ruler is used to calculate rock quality indicators, joint occurrence (including dip and inclination), joint surface density and joint surface thickness, and to observe joint surface filling and cementation characteristics and joint surface morphological characteristics.

[0071] The excavation loss collection module is used to collect energy consumption and material consumption during the TBM excavation process, including resources consumed to maintain cutterhead rotation, resources consumed to maintain cutterhead thrust, resources consumed to maintain cylinder pressure, and cutter wear;

[0072] The excavation parameter collection module is used to collect TBM excavation parameter data during the TBM excavation process, including excavation speed, excavation thrust, cutterhead speed and cutterhead thrust.

[0073] In this embodiment, the data processing module is used to process various types of data obtained in the testing and acquisition module to obtain corresponding feature vectors;

[0074] Specifically:

[0075] (1) For the TSP seismic wave test data, the longitudinal wave velocity, shear wave velocity, Poisson's ratio and density are selected according to certain rules and the TSP seismic wave characteristic vector is generated;

[0076] Specifically, along the excavation direction, 10 cm is used as a calculation unit, and the longitudinal wave velocity, shear wave velocity, Poisson's ratio and density at both ends of the calculation unit are arranged in order to form the TSP seismic wave characteristic vector.

[0077] (2) For the induced polarization test data, the resistivity is selected according to certain rules and the induced polarization characteristic vector is generated;

[0078] Specifically, along the excavation direction, 10 cm is used as a calculation unit, and the resistivity near the five drilling points is selected to represent the resistivity of the entire calculation unit; Figure 2 As shown;

[0079] The resistivity diagram of borehole No. 1 position 1 is as follows Figure 3 As shown, any two adjacent resistivity points are 5 cm apart. The resistivity values ​​within the calculation unit are arranged in order from left to right, from top to bottom, and from front to back to form the resistivity vector of borehole position 1 No. 1. The resistivity vectors of any five holes are combined in the order of the holes to form the induced polarization characteristic vector of the entire calculation unit.

[0080] (3) Based on the advance drilling test data, the joint surface filling and cementation characteristics, joint surface morphology characteristics, rock quality indicators, joint occurrence (including dip and inclination), joint surface density, and joint surface thickness are obtained, and the joint characteristic vector is obtained along the excavation direction with 10 cm as a calculation unit;

[0081] The rock characteristic matrix is ​​generated using the point load test results and the wear resistance test results. The test results are directly imported into the rock characteristic matrix along the excavation direction with 10 cm as a calculation unit.

[0082] In this embodiment, the TSP seismic wave characteristic vector, the induced polarization characteristic vector, the joint characteristic vector and the rock characteristic array are sequentially used to generate the surrounding rock characteristic vector; the tunneling loss vector and tunneling vector are obtained according to the tunneling loss and tunneling parameters.

[0083] In this embodiment, the deep learning model uses an LSTM neural network. The surrounding rock feature vectors and their corresponding loss vectors and advance vectors during the tunneling process are used to construct a training database to train the LSTM neural network. The training database is continuously expanded during the tunneling process. When the training database increases to a certain amount, the deep learning model is retrained to improve accuracy.

[0084] In this embodiment, the surrounding rock characteristic vector R includes seismic wave data, induced polarization test data, point load strength, wear resistance data, etc., and may also include fracture occurrence, fracture width, fracture density, fracture filling data, etc.;

[0085] The excavation vector includes the excavation speed v, the excavation thrust, the cutterhead speed and the excavation data T of the cutterhead thrust;

[0086] The tunneling loss vector loss includes the resources consumed to maintain the cutterhead rotation, the resources consumed to maintain the cutterhead thrust, the resources consumed to maintain the cylinder pressure, and the cutter wear;

[0087] Perform relationship modeling on R1, R2, T1, T2 at the excavated position, R3, T3 at the excavated position, and R4, R5 at the unexcavated position to solve v3 and loss3 at the excavated position, that is, (v3, loss3) = F(R1, T1, R2, T2, R3, T3, R4, R5); LSTM or GRU models can be used for relationship modeling.

[0088] like Figure 4 As shown, the intelligent decision-making module obtains the excavation loss under different excavation parameters based on the current surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, and then finds the optimal excavation vector through the optimization algorithm based on the received excavation loss range, excavation parameter range and excavation mode to provide excavation parameter recommendations.

[0089] Among them, the excavation mode includes energy-saving mode and rapid advancement mode; the energy-saving mode is to select the excavation mode with the smallest excavation loss while ensuring that the excavation speed meets the advancement speed requirements; the rapid advancement mode is to select the fastest excavation speed within the acceptable range of excavation loss.

[0090] In this embodiment, the optimization algorithm includes an energy-saving mode optimization algorithm and a fast-advance mode optimization algorithm;

[0091] The energy-saving mode optimization algorithm is to set the minimum excavation speed v min , while meeting the minimum tunneling speed v min Minimize excavation loss under the requirements; specifically:

[0092] (1) Based on the surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, the executable decision parameters are discretized to generate a number of initial decision schemes;

[0093] (2) Substitute group a data into the trained deep learning model to obtain the initial array (v, loss) of group a’s tunneling speed and tunneling loss;

[0094] (3) Select the initial array (v, loss) that meets the tunneling speed condition (v>v min ), and then select group b of tunneling parameter arrays according to the principle of minimizing tunneling loss;

[0095] (4) Randomly select three from the tunneling parameter array of group b and perform differential analysis;

[0096] D4=D1+f(D2-D3)

[0097] Where D1, D2 and D3 are the data of the excavated position, f is a constant, and D4 is the data after difference;

[0098] (5) Repeat step (4) c times to obtain C candidate decision solutions;

[0099] (6) Substitute the C candidate decision solutions into the trained deep learning model to obtain c sets of arrays (v, loss);

[0100] (7) Repeat steps (4)-(6) e times;

[0101] (8) Select the decision plan that minimizes the excavation loss from the new decision plans.

[0102] The optimization algorithm for the fast advancement mode is to set the maximum excavation loss loss max , while meeting the maximum tunneling loss loss max Under the requirement, maximize the excavation speed v; specifically:

[0103] (1) Discrete the executable decision parameters to generate a number of initial decision solutions;

[0104] (2) Substitute group a data into the trained deep learning model to obtain the initial array (v, loss) of group a’s tunneling speed and tunneling loss;

[0105] (3) Select the initial array (v, loss) from group a that meets the tunneling loss (loss <loss max ), and then select b groups of excavation parameters according to the principle of maximizing the excavation speed v;

[0106] (4) Randomly select three from the b group of excavation parameters for difference:

[0107] D4=D1+f(D2-D3)

[0108] Where D1, D2 and D3 are the data of the excavated position, f is a constant, and D4 is the data after difference;

[0109] (5) Repeat step (4) c times to obtain C candidate decision solutions;

[0110] (6) Substitute the C candidate decision solutions into the trained deep learning model to obtain c sets of arrays (v, loss);

[0111] (7) Repeat steps (4)-(6) e times;

[0112] (8) Select the decision plan corresponding to the maximum excavation speed v from the new decision plans.

[0113] Example 2

[0114] like Figure 5 As shown, this embodiment provides a TBM intelligent decision-making method based on deep learning, including the following steps:

[0115] Conduct tunneling tests on the surrounding rock in front of the tunnel face and obtain seismic wave test data, induced polarization test data, advance drilling test data, tunneling losses, and tunneling parameters;

[0116] The surrounding rock characteristic vector is obtained based on the seismic wave test data, the induced polarization test data and the advance drilling test data, and the excavation loss vector and excavation vector are obtained based on the excavation loss and excavation parameters respectively;

[0117] Receive preset excavation loss range, excavation parameter range and excavation mode;

[0118] According to the surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, the trained deep learning model is used to obtain the excavation loss under different excavation parameters. The excavation parameters are selected according to the excavation loss range and excavation mode.

[0119] The TBM intelligent decision-making system based on deep learning described in Example 1 is used, and the specific implementation method is as follows:

[0120] A. Start the induced polarization instrument to test the surrounding rock in front of the tunnel face;

[0121] B. Start the TSP seismic wave tester to test the surrounding rock in front of the tunnel face;

[0122] C. Use the drilling rig to drill holes at the tunnel face, record while drilling, and judge and record the joint occurrence, joint thickness, and joint characteristics through the drill core;

[0123] D. Conduct point load test and wear resistance test on the drilled core;

[0124] E. Obtain the seismic wave test data, induced polarization test data, and advance drilling test data obtained in the above test process; generate corresponding vectors for the relevant parameters along the excavation direction with 10 cm as a calculation unit, and integrate them to form the excavation vector, loss vector, and surrounding rock feature vector;

[0125] F. Select the tunneling mode based on the shield machine conditions, material preparation, preset acceptable tunneling loss range and tunneling parameter range;

[0126] G. Use the trained deep learning model of the mapping relationship between surrounding rock characteristics and tunneling parameters to predict tunneling losses under different tunneling conditions and select tunneling parameters based on the tunneling mode;

[0127] H. Continuously collect tunneling parameters during TBM operation and check the TBM during tunneling step changes to obtain TBM machine wear and tear information, so as to continuously train and update the model and improve the model prediction accuracy.

[0128] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A TBM intelligent decision-making system based on deep learning, characterized by: include: The testing and acquisition module is used to conduct tunneling tests on the surrounding rock in front of the tunnel face and obtain seismic wave test data, induced polarization test data, advance drilling test data, tunneling losses, and tunneling parameters; a data set processing module configured to obtain a surrounding rock feature vector based on seismic wave test data, induced polarization test data, and advance drilling test data, and to obtain an excavation loss vector and an excavation vector based on excavation loss and excavation parameters, respectively; An intelligent decision-making module receives a preset excavation loss range, excavation parameter range, and excavation mode, and is configured to use a trained deep learning model to determine the excavation loss under different excavation parameters based on the surrounding rock feature vector and the corresponding excavation loss vector and excavation vector, and select the excavation parameters based on the excavation loss range and excavation mode; The excavation mode includes an energy-saving mode, which is a excavation mode that minimizes excavation loss while ensuring that the excavation speed meets the advancement speed requirement; In energy-saving mode, the optimal tunneling vector is selected through an optimization algorithm, and the minimum tunneling speed is set to minimize tunneling losses while meeting the minimum tunneling speed requirement. Specifically, (1) Generate an initial decision plan; (2) Applying the trained deep learning model to the initial decision-making scheme to obtain the initial array of tunneling speed and tunneling loss; (3) Filter out the array that meets the minimum excavation speed requirement from the initial array, and then filter out the corresponding excavation parameter array based on the excavation loss; (4) Randomly select any three data from the tunneling parameter array and perform differential analysis repeatedly to obtain a candidate decision solution; (5) Applying the trained deep learning model to the candidate decision-making schemes to obtain an array of tunneling speed and tunneling loss; (6) Repeat steps (4)-(5) and select the decision plan that corresponds to the minimum excavation loss from the new decision plans obtained.

2. The TBM intelligent decision-making system based on deep learning according to claim 1, characterized in that: The testing and acquisition module includes a seismic wave testing module, which is used to perform advance detection of the surrounding rock in front of the tunnel face to obtain the longitudinal wave velocity, shear wave velocity, Poisson's ratio and density, and thereby obtain the seismic wave characteristic vector.

3. The TBM intelligent decision-making system based on deep learning according to claim 1, characterized in that: The testing and acquisition module includes an induced polarization testing module, which is used to detect the spatial distribution characteristics of the surrounding rock resistivity and obtain the induced polarization characteristic vector.

4. The TBM intelligent decision-making system based on deep learning according to claim 1, characterized in that: The testing and acquisition module includes an advance drilling testing module; the advance drilling testing module drills and samples the surrounding rock in front of the tunnel face to obtain rock quality indicators, joint occurrence, joint surface density, joint surface thickness, joint surface filling and cementation characteristics and joint surface morphological characteristics, and thereby obtains a joint characteristic vector.

5. The TBM intelligent decision-making system based on deep learning according to claim 4, characterized in that: The advance drilling test module further includes performing a point load test and a friction test on the surrounding rock in front of the tunnel face, and obtaining a rock characteristic matrix based on the point load test results and the wear resistance test results.

6. The TBM intelligent decision-making system based on deep learning according to claim 1, characterized in that: The testing and collection module includes an excavation loss collection module, which is used to collect resources consumed in maintaining cutterhead rotation, resources consumed in maintaining cutterhead thrust, resources consumed in maintaining oil cylinder pressure, and cutter wear.

7. The TBM intelligent decision-making system based on deep learning according to claim 1, characterized in that: The testing and collection module includes a tunneling parameter collection module, which is used to collect tunneling speed, tunneling thrust, cutterhead rotation speed and cutterhead thrust.

8. The TBM intelligent decision-making system based on deep learning according to claim 1, characterized in that: The excavation mode also includes a fast advance mode, which is a excavation mode that selects the fastest excavation speed when the excavation loss is within an acceptable range. In the fast advance mode, the optimal excavation vector is selected through an optimization algorithm; Set the maximum excavation loss and maximize the excavation speed while meeting the maximum excavation loss requirement; Specifically include: (1) Generate an initial decision plan; (2) Applying the trained deep learning model to the initial decision-making scheme to obtain the initial array of tunneling speed and tunneling loss; (3) Filter out the array that meets the tunneling loss requirements from the initial array, and then filter out the corresponding tunneling parameter array according to the tunneling speed; (4) Randomly select any three data from the tunneling parameter array and perform differential analysis repeatedly to obtain a candidate decision solution; (5) Applying the trained deep learning model to the candidate decision-making schemes to obtain an array of tunneling speed and tunneling loss; (6) Repeat steps (4)-(5) and select the decision plan corresponding to the maximum excavation speed from the new decision plans obtained.

9. A TBM intelligent decision-making method based on deep learning, characterized in that: include: Conduct tunneling tests on the surrounding rock in front of the tunnel face and obtain seismic wave test data, induced polarization test data, advance drilling test data, tunneling losses, and tunneling parameters; The surrounding rock characteristic vector is obtained based on the seismic wave test data, the induced polarization test data and the advance drilling test data, and the excavation loss vector and excavation vector are obtained based on the excavation loss and excavation parameters respectively; Receive preset excavation loss range, excavation parameter range and excavation mode; Based on the surrounding rock feature vector and the corresponding tunneling loss vector and tunneling vector, the trained deep learning model is used to obtain the tunneling loss under different tunneling parameters. The tunneling parameters are selected according to the tunneling loss range and tunneling mode. The excavation mode includes an energy-saving mode, which is a excavation mode that minimizes excavation loss while ensuring that the excavation speed meets the advancement speed requirement; In energy-saving mode, the optimal tunneling vector is selected through an optimization algorithm, and the minimum tunneling speed is set to minimize tunneling losses while meeting the minimum tunneling speed requirement. Specifically, (1) Generate an initial decision plan; (2) Applying the trained deep learning model to the initial decision-making scheme to obtain the initial array of tunneling speed and tunneling loss; (3) Filter out the array that meets the minimum excavation speed requirement from the initial array, and then filter out the corresponding excavation parameter array based on the excavation loss; (4) Randomly select any three data from the tunneling parameter array and perform differential analysis repeatedly to obtain a candidate decision solution; (5) Applying the trained deep learning model to the candidate decision-making schemes to obtain an array of tunneling speed and tunneling loss; (6) Repeat steps (4)-(5) and select the decision plan that corresponds to the minimum excavation loss from the new decision plans obtained.

Images