Surface convergence settlement prediction model training method, system and application method, system

Abstract

The present invention discloses a surface convergence settlement prediction model training method, system and application method and system, which relate to the field of surface settlement prediction. The present invention first trains an auxiliary model and a main model to obtain a trained auxiliary model and a main model. The complete shield tunneling parameters can be obtained through the auxiliary model, thereby ensuring the integrity of the shield machine technical parameters for advance prediction of surface convergence settlement of shield construction. Compared with the prior art, the present invention solves the problem of incomplete shield machine technical parameters for advance prediction of surface convergence settlement of shield construction through the auxiliary model, and realizes high-precision prediction of surface settlement of shield construction through the complete shield machine technical parameters and the main model.

Description

Technical Field

[0001] The present invention relates to the field of ground surface settlement prediction for subway construction, and in particular to a ground surface convergence settlement prediction model training method and system and an application method and system. Background Art

[0002] During the construction of urban shield tunnels, the ground disturbance caused by shield construction will inevitably generate additional loads on the adjacent soil and existing structures, which will then cause settlement and deformation. When the maximum surface deformation induced by subway construction exceeds the allowable value of the existing structure, it will affect its normal service and even cause engineering accidents. Therefore, it is necessary to make advance predictions of the surface convergence and settlement caused by shield construction. At present, a large number of studies have proposed the use of intelligent methods to achieve advance prediction of the surface during shield construction. However, since the parameters of the shield machine between the shield machine face and the prediction point are partially unknown, a large number of intelligent methods currently do not consider the technical parameters of the shield machine. The advance prediction of surface convergence and settlement that abandons the technical parameters of the shield machine has poor prediction effect and low model accuracy due to the lack of input parameters. Summary of the invention

[0003] The purpose of the present invention is to provide a surface convergence settlement prediction model training method, system and application method and system, which can achieve high-precision shield construction surface settlement prediction.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A surface convergence settlement prediction model training method, the training method comprising the following steps:

[0006] Constructing a surface convergence settlement prediction model based on a neural network, the surface convergence settlement prediction model includes an auxiliary model and a main model, the auxiliary model is used to predict the response parameters of the shield machine corresponding to the target position, and the main model is used to predict the surface convergence settlement amount corresponding to the target position;

[0007] Acquire engineering data corresponding to each excavation layer, wherein the excavation layer is a construction layer corresponding to the tunnel face of the shield machine; the engineering data includes: geometric parameters, shield tail grouting parameters, shield excavation parameters, geological parameters and surface convergence settlement, and the shield excavation parameters include: active control parameters of the shield machine and response parameters of the shield machine;

[0008] Inputting active control parameters of the shield machine corresponding to the first excavation layer and geological parameters corresponding to the second excavation layer into the auxiliary model; the first excavation layer is any layer among all excavation layers, and the second excavation layer is any layer after the first excavation layer;

[0009] Determining a first loss function according to the output of the auxiliary model and the response parameter of the shield machine corresponding to the second excavation layer;

[0010] Optimizing the parameters in the auxiliary model according to the first loss function to train the auxiliary model to obtain a trained auxiliary model;

[0011] Inputting geometric parameters corresponding to the first excavation layer, shield tail grouting parameters corresponding to the first excavation layer, shield excavation parameters corresponding to the second excavation layer, and geological parameters corresponding to the second excavation layer into the main model;

[0012] determining a second loss function according to the output of the main model and the surface convergence settlement corresponding to the second excavation layer;

[0013] Optimize the parameters in the main model according to the second loss function to train the main model and obtain a trained main model.

[0014] Optionally, the training method further includes:

[0015] The engineering data corresponding to each excavation layer is aligned to obtain processed data.

[0016] Optionally, the geometric parameters include tunnel diameter, tunnel burial depth, and lining thickness; the shield tail grouting parameters include grouting pressure, grouting volume, and shield tail gap; the shield tunneling parameters include soil compartment pressure, total thrust, propulsion speed, cutter head torque, and cutter head speed; the geological parameters include soil layer thickness, soil type, groundwater level, and mechanical parameters.

[0017] The present invention also provides a surface convergence settlement prediction model application method, the application method comprising the following steps:

[0018] Obtain geological parameters corresponding to the target position, geometric parameters corresponding to the current excavation layer, shield tail grouting parameters corresponding to the current excavation layer, and active control parameters of the shield machine corresponding to the current excavation layer;

[0019] Inputting the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into the auxiliary model to obtain the response parameters of the shield machine corresponding to the target position, wherein the auxiliary model is a model trained according to the surface convergence settlement prediction model training method described above;

[0020] Determine the shield tunneling parameters corresponding to the target position according to the active control parameters of the shield machine corresponding to the current tunneling layer and the response parameters of the shield machine corresponding to the target position;

[0021] The geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position are input into the main model to obtain the surface convergence settlement corresponding to the target position. The main model is a model trained according to the above-mentioned surface convergence settlement prediction model training method.

[0022] The present invention also provides a surface convergence settlement prediction model training system, the training system comprising:

[0023] A model building unit, used to build a surface convergence settlement prediction model based on a neural network, wherein the surface convergence settlement prediction model includes an auxiliary model and a main model, wherein the auxiliary model is used to predict the response parameters of the shield machine corresponding to the target position, and the main model is used to predict the surface convergence settlement amount corresponding to the target position;

[0024] An engineering data acquisition unit is used to acquire engineering data corresponding to each excavation layer, wherein the excavation layer is a construction layer corresponding to the tunnel face of the shield machine; the engineering data includes: geometric parameters, shield tail grouting parameters, shield excavation parameters, geological parameters and surface convergence settlement, and the shield excavation parameters include: active control parameters of the shield machine and response parameters of the shield machine;

[0025] A first input unit is used to input active control parameters of the shield machine corresponding to a first excavation layer and geological parameters corresponding to a second excavation layer into the auxiliary model; the first excavation layer is any layer among all excavation layers, and the second excavation layer is any layer after the first excavation layer;

[0026] A first loss function determination unit, configured to determine a first loss function according to an output of the auxiliary model and a response parameter of the shield machine corresponding to the second excavation layer;

[0027] A first training unit, used for optimizing the parameters in the auxiliary model according to the first loss function, so as to train the auxiliary model and obtain a trained auxiliary model;

[0028] A second input unit is used to input geometric parameters corresponding to the first excavation layer, shield tail grouting parameters corresponding to the first excavation layer, shield excavation parameters corresponding to the second excavation layer, and geological parameters corresponding to the second excavation layer into the main model;

[0029] a second loss function determination unit, configured to determine a second loss function according to an output of the main model and a surface convergence settlement amount corresponding to the second excavation layer;

[0030] The second training unit is used to optimize the parameters in the main model according to the second loss function to train the main model to obtain a trained main model.

[0031] Optionally, the training system further includes:

[0032] The alignment processing unit is used to perform alignment processing on the engineering data corresponding to each excavation layer to obtain processed data.

[0033] Optionally, the geometric parameters include tunnel diameter, tunnel burial depth, and lining thickness; the shield tail grouting parameters include grouting pressure, grouting volume, and shield tail gap; the shield tunneling parameters include soil compartment pressure, total thrust, propulsion speed, cutter head torque, and cutter head speed; the geological parameters include soil layer thickness, soil type, groundwater level, and mechanical parameters.

[0034] The present invention also provides a surface convergence settlement prediction model application system, the application system comprising:

[0035] A parameter acquisition unit, used to acquire geological parameters corresponding to the target position, geometric parameters corresponding to the current excavation layer, shield tail grouting parameters corresponding to the current excavation layer, and active control parameters of the shield machine corresponding to the current excavation layer;

[0036] A response parameter prediction unit of a shield machine, used to input the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into an auxiliary model to obtain the response parameters of the shield machine corresponding to the target position, wherein the auxiliary model is a model trained according to the surface convergence settlement prediction model training method described above;

[0037] A shield tunneling parameter determination unit, used to determine the shield tunneling parameters corresponding to the target position according to the active control parameters of the shield machine corresponding to the current tunneling layer and the response parameters of the shield machine corresponding to the target position;

[0038] The surface convergence and settlement prediction unit is used to input the geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position into the main model to obtain the surface convergence and settlement corresponding to the target position. The main model is a model trained according to the surface convergence and settlement prediction model training method described above.

[0039] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0040] The present invention provides a surface convergence settlement prediction model training method, system and application method and system. In the present invention, engineering data corresponding to each excavation layer is obtained, and an auxiliary model is trained according to the active control parameters of the shield machine corresponding to the first excavation layer and the geological parameters corresponding to the second excavation layer, wherein the first excavation layer is any layer among all excavation layers, and the second excavation layer is any layer after the first excavation layer; the main model is trained according to the geometric parameters corresponding to the first excavation layer, the shield tail grouting parameters corresponding to the first excavation layer, the shield excavation parameters corresponding to the second excavation layer, and the geological parameters corresponding to the second excavation layer ; Input the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into the trained auxiliary model, and the response parameters of the shield machine corresponding to the target position can be obtained; According to the active control parameters of the shield machine corresponding to the current excavation layer and the response parameters of the shield machine corresponding to the target position, the shield excavation parameters corresponding to the target position can be determined; The geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position are input into the main model, and the surface convergence settlement corresponding to the target position can be obtained. The present invention can obtain complete shield excavation parameters through the auxiliary model, and then ensure the integrity of the shield machine technical parameters for advance prediction of surface convergence settlement of shield construction. Compared with the prior art, the present invention solves the problem of incomplete shield machine technical parameters for advance prediction of surface convergence settlement of shield construction through the auxiliary model, and realizes high-precision prediction of surface settlement of shield construction through complete shield machine technical parameters and the main model. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 A flow chart of a surface convergence settlement prediction model training method provided in Example 1 of the present invention;

[0043] Figure 2 It is the surface convergence settlement prediction diagram for shield construction;

[0044] Figure 3 A flow chart of a method for applying a surface convergence settlement prediction model provided in Example 2 of the present invention;

[0045] Figure 4 This is the main model structure diagram;

[0046] Figure 5A structural block diagram of a surface convergence settlement prediction model training system provided in Example 3 of the present invention;

[0047] Figure 6 This is a structural block diagram of a surface convergence and settlement prediction model application system provided in Example 4 of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] During the construction of urban shield tunnels, the ground disturbance caused by shield construction will inevitably generate additional loads on the adjacent soil and existing structures, which will in turn cause settlement and deformation. When the deformation induced by the construction exceeds the allowable value of the existing structure, it will affect its normal service and even cause engineering accidents. Therefore, it is necessary to establish an advance prediction of the surface settlement caused by shield construction in the project. At present, there is no theoretical solution for the plasticity of shield three-dimensional construction; numerical simulation methods generally have poor prediction results because they require the input of certain and detailed soil parameters; although model tests are highly accurate, they are too expensive and cannot be used on a large scale. At present, a large number of studies have proposed the use of intelligent methods to achieve advanced prediction of the ground surface during shield construction. However, since some of the shield machine parameters between the shield machine face and the predicted point are unknown (such as Figure 2 As shown in Figure 1, currently, a large number of intelligent methods can only predict the settlement above the shield machine face or the surface convergence settlement advance prediction without considering the shield machine technical parameters. The settlement prediction above the shield machine face is too late, and there is not enough time for the construction personnel to take auxiliary control measures, so it is not very practical in the project. The surface convergence settlement advance prediction without the shield machine technical parameters has poor prediction effect and low model accuracy due to the lack of input parameters.

[0050] The purpose of the present invention is to provide a surface convergence settlement prediction model training method, system and application method and system, which can achieve high-precision shield construction surface settlement prediction.

[0051] Terminology explanation:

[0052] Shield construction: When constructing subway tunnels, for sections that are mainly soil, shield machines are used for subway tunnel construction.

[0053] Stratum loss: In underground tunnels, over-excavation and soil deformation will inevitably occur. These factors cause stratum displacement, which are collectively referred to as stratum loss.

[0054] Shield construction settlement: Shield construction will inevitably cause ground disturbance, and when the ground disturbance further develops to the surface, surface settlement will occur.

[0055] Active control parameters of the shield machine: During the construction of the earth pressure balance shield machine, the shield machine driver mainly controls the shield machine parameters by controlling the shield machine's propulsion speed and cutter head speed to stabilize within a certain numerical range. Therefore, the shield machine's soil compartment pressure, cutter head speed and propulsion speed can be used as the shield machine's hole control parameters.

[0056] Response parameters of the shield machine: When the shield machine controls the propulsion speed, cutter head speed and soil chamber pressure to excavate in different strata, the cutter head torque and total thrust of the shield machine will vary due to different strata, which are the response parameters of the shield machine.

[0057] Advance prediction of surface convergence and settlement: During tunnel construction, when the shield machine face is still some distance away from the predicted point, the surface settlement can be predicted.

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] Embodiment 1:

[0060] See also Figure 1 The present invention provides a surface convergence settlement prediction model training method, the training method comprises the following steps:

[0061] A1: A surface convergence settlement prediction model is constructed based on a neural network. The surface convergence settlement prediction model includes an auxiliary model and a main model. The auxiliary model is used to predict the response parameters of the shield machine corresponding to the target position, and the main model is used to predict the surface convergence settlement amount corresponding to the target position;

[0062] A2: Obtain the engineering data corresponding to each excavation layer, where the excavation layer is the construction layer corresponding to the shield machine face; the engineering data include: geometric parameters, shield tail grouting parameters, shield excavation parameters, geological parameters and surface convergence settlement, and the shield excavation parameters include: active control parameters of the shield machine and response parameters of the shield machine; wherein the geometric parameters include tunnel diameter, tunnel burial depth and lining thickness, representing the basic characteristics of the project; the shield tail grouting parameters include grouting pressure, grouting volume and shield tail gap; the shield excavation parameters include soil cabin pressure, total thrust, propulsion speed, cutter head torque and cutter head speed; the geological parameters include soil layer thickness, soil type, groundwater level and mechanical parameters, which can be obtained from geological exploration.

[0063] A3: inputting active control parameters of the shield machine corresponding to the first excavation layer and geological parameters corresponding to the second excavation layer into the auxiliary model; the first excavation layer is any layer among all excavation layers, and the second excavation layer is any layer after the first excavation layer;

[0064] It should be noted that the engineering data corresponding to each excavation layer are obtained according to the order of advancement of the shield machine. The second excavation layer is any layer after the first excavation layer, which means that the engineering data corresponding to the second excavation layer is obtained after the engineering data corresponding to the first excavation layer is obtained.

[0065] A4: determining a first loss function according to the output of the auxiliary model and the response parameter of the shield machine corresponding to the second tunneling layer;

[0066] A5: Optimizing the parameters in the auxiliary model according to the first loss function to train the auxiliary model to obtain a trained auxiliary model;

[0067] A6: inputting geometric parameters corresponding to the first excavation layer, shield tail grouting parameters corresponding to the first excavation layer, shield excavation parameters corresponding to the second excavation layer, and geological parameters corresponding to the second excavation layer into the main model;

[0068] A7: determining a second loss function according to the output of the main model and the surface convergence settlement corresponding to the second excavation layer;

[0069] A8: Optimize the parameters in the main model according to the second loss function to train the main model and obtain a trained main model.

[0070] As a possible implementation manner, the training method further includes:

[0071] The engineering data corresponding to each excavation layer is aligned to obtain processed data.

[0072] When training the auxiliary model, the data set including the active control parameters and geological parameters of the shield machine is divided into a training set and a test set, with a ratio of 8.5:1.5. First, the auxiliary model is trained with the training set. During the training, the loss function of the auxiliary model on the training set needs to be calculated. As the training progresses, the loss function of the auxiliary model for the training set will gradually decrease, and the fitting effect will continue to improve. The noise in the data may also be learned into the model, resulting in poor fitting effect on the test set, which is called "overfitting". In order to prevent the auxiliary model from overfitting, the training needs to be terminated in advance. Finally, after the auxiliary model is trained, the auxiliary model is tested using the test set.

[0073] Similarly, when training the main model, the data set including geometric parameters, shield tail grouting parameters, shield tunneling parameters and geological parameters is divided into a training set and a test set, with a ratio of 8.5:1.5. First, the main model is trained with the training set. During the training, the loss function of the main model on the training set needs to be calculated. As the training progresses, the loss function of the main model for the training set will gradually decrease, and the fitting effect will continue to improve. The noise in the data may also be learned into the model, resulting in poor fitting effect on the test set, which is called "overfitting". In order to prevent the main model from overfitting, the training needs to be terminated in advance. Finally, the main model is trained and tested with the test set.

[0074] The response parameters of the shield machine corresponding to the target position can be predicted by the trained auxiliary model. According to the active control parameters of the shield machine corresponding to the current excavation layer and the response parameters of the shield machine corresponding to the target position, the shield excavation parameters corresponding to the target position can be determined, that is, the complete shield excavation parameters can be obtained. The surface convergence settlement corresponding to the target position can be predicted through the complete shield excavation parameters, geometric parameters, geological parameters and shield tail grouting parameters. Since the present invention takes into account the problem of incomplete technical parameters of the shield machine, high-precision prediction of surface settlement of shield construction can be achieved.

[0075] Embodiment 2:

[0076] See also Figure 3 The present invention provides a surface convergence settlement prediction model application method, the application method comprising the following steps:

[0077] B1: Obtain geological parameters corresponding to the target position, geometric parameters corresponding to the current excavation layer, shield tail grouting parameters corresponding to the current excavation layer, and active control parameters of the shield machine corresponding to the current excavation layer;

[0078] B2: Inputting the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into the auxiliary model to obtain the response parameters of the shield machine corresponding to the target position, wherein the auxiliary model is a model trained according to the surface convergence settlement prediction model training method described in Example 1;

[0079] B3: Determine the shield tunneling parameters corresponding to the target position according to the active control parameters of the shield machine corresponding to the current tunneling layer and the response parameters of the shield machine corresponding to the target position;

[0080] B4: The geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position are input into the main model to obtain the surface convergence settlement corresponding to the target position. The main model is a model trained according to the surface convergence settlement prediction model training method described in Example 1.

[0081] In practical applications, such as Figure 4 As shown in the figure, after completing the shield tunneling parameters based on the auxiliary model, complete geometric parameters, geological parameters, shield tunneling parameters and shield tail grouting parameters can be obtained. Then, each parameter is encoded separately, such as the geometric parameters can be encoded as a 1×6 array, the geological parameters can be encoded as a 5×20×2 array, where 5 represents the number of tunneling layers obtained, and the shield machine tunneling parameters will form a 5×15 array, where 15 represents the number of tunneling layers obtained, and the shield tail grouting parameters will form a 3×8 array, where 8 represents the number of tunneling layers obtained. The geological parameter array is first reduced to a 1×5×2 array using CNN (convolutional neural network), and then formed into a 1×3 array using DNN (dense neural network). For the shield tunneling parameter array, LSTM (long short-term memory network) is first used to form a 1×15 array, and then DNN (dense neural network) is used to form a 1×3 array. For the shield tail grouting parameter array, Bi-LSTM (bidirectional long short-term memory network) is first used to form a 1×8 array, and then DNN (dense neural network) is used to form a 1×3 array. The unprocessed geometric parameter array is connected with the processed geological parameter array, shield tunneling array and shield tail grouting array to form a 1×15 array, and finally DNN (dense neural network) is used to output the surface convergence settlement.

[0082] The present invention can obtain complete shield machine technical parameters through auxiliary models, thereby ensuring the integrity of shield machine technical parameters for advance prediction of surface convergence and settlement during shield construction. Compared with the prior art, the present invention solves the problem of incomplete shield machine technical parameters for advance prediction of surface convergence and settlement during shield construction through auxiliary models, and realizes high-precision prediction of surface settlement during shield construction through complete shield machine technical parameters and main models.

[0083] Embodiment 3:

[0084] See also Figure 5 The present invention provides a surface convergence settlement prediction model training system, the training system comprising:

[0085] A model building unit M1 is used to build a surface convergence settlement prediction model based on a neural network, wherein the surface convergence settlement prediction model includes an auxiliary model and a main model, wherein the auxiliary model is used to predict the response parameters of the shield machine corresponding to the target position, and the main model is used to predict the surface convergence settlement amount corresponding to the target position;

[0086] The engineering data acquisition unit M2 is used to obtain the engineering data corresponding to each excavation layer, and the excavation layer is the construction layer corresponding to the shield machine face; the engineering data includes: geometric parameters, shield tail grouting parameters, shield excavation parameters, geological parameters and surface convergence settlement, and the shield excavation parameters include: active control parameters of the shield machine and response parameters of the shield machine; wherein, the geometric parameters include tunnel diameter, tunnel burial depth, and lining thickness; the shield tail grouting parameters include grouting pressure, grouting volume and shield tail gap; the shield excavation parameters include soil cabin pressure, total thrust, propulsion speed, cutter head torque and cutter head speed; the geological parameters include soil layer thickness, soil type, groundwater level and mechanical parameters.

[0087] The first input unit M3 is used to input the active control parameters of the shield machine corresponding to the first excavation layer and the geological parameters corresponding to the second excavation layer into the auxiliary model; the first excavation layer is any layer among all the excavation layers, and the second excavation layer is any layer after the first excavation layer;

[0088] A first loss function determination unit M4, configured to determine a first loss function according to an output of the auxiliary model and a response parameter of the shield machine corresponding to the second excavation layer;

[0089] A first training unit M5, configured to optimize the parameters in the auxiliary model according to the first loss function, so as to train the auxiliary model and obtain a trained auxiliary model;

[0090] A second input unit M6 is used to input geometric parameters corresponding to the first excavation layer, shield tail grouting parameters corresponding to the first excavation layer, shield excavation parameters corresponding to the second excavation layer, and geological parameters corresponding to the second excavation layer into the main model;

[0091] A second loss function determination unit M7, configured to determine a second loss function according to the output of the main model and the surface convergence settlement amount corresponding to the second excavation layer;

[0092] The second training unit M8 is used to optimize the parameters in the main model according to the second loss function to train the main model to obtain a trained main model.

[0093] Specifically, the training system further includes:

[0094] The alignment processing unit is used to perform alignment processing on the engineering data corresponding to each excavation layer to obtain processed data.

[0095] Embodiment 4:

[0096] See also Figure 6 The present invention provides a surface convergence settlement prediction model application system, the application system comprising:

[0097] The parameter acquisition unit N1 is used to acquire the geological parameters corresponding to the target position, the geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, and the active control parameters of the shield machine corresponding to the current excavation layer;

[0098] The response parameter prediction unit N2 of the shield machine is used to input the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into the auxiliary model to obtain the response parameters of the shield machine corresponding to the target position, and the auxiliary model is a model trained according to the surface convergence settlement prediction model training method described in Example 1;

[0099] The shield tunneling parameter determination unit N3 is used to determine the shield tunneling parameters corresponding to the target position according to the active control parameters of the shield machine corresponding to the current tunneling layer and the response parameters of the shield machine corresponding to the target position;

[0100] The surface convergence and settlement prediction unit N4 is used to input the geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position into the main model to obtain the surface convergence and settlement corresponding to the target position. The main model is a model trained according to the surface convergence and settlement prediction model training method described in Example 1.

[0101] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0102] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A surface convergence settlement prediction model training method, characterized in that: The training method comprises the following steps: Constructing a surface convergence settlement prediction model based on a neural network, the surface convergence settlement prediction model includes an auxiliary model and a main model, the auxiliary model is used to predict the response parameters of the shield machine corresponding to the target position, and the main model is used to predict the surface convergence settlement amount corresponding to the target position; Acquire engineering data corresponding to each excavation layer, wherein the excavation layer is a construction layer corresponding to the tunnel face of the shield machine; the engineering data includes: geometric parameters, shield tail grouting parameters, shield excavation parameters, geological parameters and surface convergence settlement, and the shield excavation parameters include: active control parameters of the shield machine and response parameters of the shield machine; Inputting active control parameters of the shield machine corresponding to the first excavation layer and geological parameters corresponding to the second excavation layer into the auxiliary model; the first excavation layer is any layer among all excavation layers, and the second excavation layer is any layer after the first excavation layer; Determining a first loss function according to the output of the auxiliary model and the response parameter of the shield machine corresponding to the second excavation layer; Optimizing the parameters in the auxiliary model according to the first loss function to train the auxiliary model and obtain a trained auxiliary model; based on the auxiliary model, shield tunneling parameters can be completed, and then complete geometric parameters, geological parameters, shield tunneling parameters and shield tail grouting parameters can be obtained; Inputting geometric parameters corresponding to the first excavation layer, shield tail grouting parameters corresponding to the first excavation layer, shield excavation parameters corresponding to the second excavation layer, and geological parameters corresponding to the second excavation layer into the main model; determining a second loss function according to the output of the main model and the surface convergence settlement corresponding to the second excavation layer; Optimize the parameters in the main model according to the second loss function to train the main model and obtain a trained main model.

2. The surface convergence settlement prediction model training method according to claim 1, characterized in that: The training method further comprises: The engineering data corresponding to each excavation layer is aligned to obtain processed data.

3. The surface convergence settlement prediction model training method according to claim 1, characterized in that: The geometric parameters include tunnel diameter, tunnel burial depth, and lining thickness; the shield tail grouting parameters include grouting pressure, grouting volume, and shield tail gap; the shield tunneling parameters include soil compartment pressure, total thrust, propulsion speed, cutter head torque, and cutter head speed; the geological parameters include soil layer thickness, soil type, groundwater level, and mechanical parameters.

4. A surface convergence settlement prediction model application method, characterized in that: The application method comprises the following steps: Obtain geological parameters corresponding to the target position, geometric parameters corresponding to the current excavation layer, shield tail grouting parameters corresponding to the current excavation layer, and active control parameters of the shield machine corresponding to the current excavation layer; Inputting the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into the auxiliary model to obtain the response parameters of the shield machine corresponding to the target position, wherein the auxiliary model is a model trained by the surface convergence settlement prediction model training method according to claim 1; Determine the shield tunneling parameters corresponding to the target position according to the active control parameters of the shield machine corresponding to the current tunneling layer and the response parameters of the shield machine corresponding to the target position; The geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position are input into the main model to obtain the surface convergence settlement corresponding to the target position. The main model is a model trained by the surface convergence settlement prediction model training method according to claim 1.

5. A surface convergence settlement prediction model training system, characterized in that: The training system comprises: A model building unit, used to build a surface convergence settlement prediction model based on a neural network, wherein the surface convergence settlement prediction model includes an auxiliary model and a main model, wherein the auxiliary model is used to predict the response parameters of the shield machine corresponding to the target position, and the main model is used to predict the surface convergence settlement amount corresponding to the target position; An engineering data acquisition unit is used to acquire engineering data corresponding to each excavation layer, wherein the excavation layer is a construction layer corresponding to the tunnel face of the shield machine; the engineering data includes: geometric parameters, shield tail grouting parameters, shield excavation parameters, geological parameters and surface convergence settlement, and the shield excavation parameters include: active control parameters of the shield machine and response parameters of the shield machine; A first input unit is used to input active control parameters of the shield machine corresponding to a first excavation layer and geological parameters corresponding to a second excavation layer into the auxiliary model; the first excavation layer is any layer among all excavation layers, and the second excavation layer is any layer after the first excavation layer; A first loss function determination unit, configured to determine a first loss function according to an output of the auxiliary model and a response parameter of the shield machine corresponding to the second excavation layer; A first training unit is used to optimize the parameters in the auxiliary model according to the first loss function to train the auxiliary model to obtain a trained auxiliary model; based on the auxiliary model, shield tunneling parameters can be completed, and then complete geometric parameters, geological parameters, shield tunneling parameters and shield tail grouting parameters can be obtained; A second input unit is used to input geometric parameters corresponding to the first excavation layer, shield tail grouting parameters corresponding to the first excavation layer, shield excavation parameters corresponding to the second excavation layer, and geological parameters corresponding to the second excavation layer into the main model; a second loss function determination unit, configured to determine a second loss function according to an output of the main model and a surface convergence settlement amount corresponding to the second excavation layer; The second training unit is used to optimize the parameters in the main model according to the second loss function to train the main model to obtain a trained main model.

6. The surface convergence settlement prediction model training system according to claim 5, characterized in that: The training system also includes: The alignment processing unit is used to perform alignment processing on the engineering data corresponding to each excavation layer to obtain processed data.

7. The surface convergence settlement prediction model training system according to claim 5, characterized in that: The geometric parameters include tunnel diameter, tunnel burial depth, and lining thickness; the shield tail grouting parameters include grouting pressure, grouting volume, and shield tail gap; the shield tunneling parameters include soil compartment pressure, total thrust, propulsion speed, cutter head torque, and cutter head speed; the geological parameters include soil layer thickness, soil type, groundwater level, and mechanical parameters.

8. A surface convergence settlement prediction model application system, characterized in that: The application system comprises: A parameter acquisition unit, used to acquire geological parameters corresponding to the target position, geometric parameters corresponding to the current excavation layer, shield tail grouting parameters corresponding to the current excavation layer, and active control parameters of the shield machine corresponding to the current excavation layer; A response parameter prediction unit of a shield machine, used for inputting the geological parameters corresponding to the target position and the active control parameters of the shield machine corresponding to the current excavation layer into an auxiliary model to obtain the response parameters of the shield machine corresponding to the target position, wherein the auxiliary model is a model trained according to the surface convergence settlement prediction model training method according to claim 1; A shield tunneling parameter determination unit, used to determine the shield tunneling parameters corresponding to the target position according to the active control parameters of the shield machine corresponding to the current tunneling layer and the response parameters of the shield machine corresponding to the target position; The surface convergence and settlement prediction unit is used to input the geometric parameters corresponding to the current excavation layer, the shield tail grouting parameters corresponding to the current excavation layer, the geological parameters corresponding to the target position and the shield excavation parameters corresponding to the target position into the main model to obtain the surface convergence and settlement corresponding to the target position. The main model is a model trained according to the surface convergence and settlement prediction model training method according to claim 1.

Images