A method for real-time prediction and adjustment of earth chamber pressure of a shield tunneling machine based on transfer learning

Abstract

The application discloses a kind of based on migration learning's shield machine soil bin pressure real-time prediction adjustment method, comprising: S1, judge whether there is the source domain task prediction model of trained good;S2, collection shield machine data set, constructs training sample set;S3, constructs economic benefit evaluation index W;S4, with the training sample set in step S2 to the neural network prediction model of pre-construction is trained, obtains initial soil bin pressure prediction model;S5, using shield machine field construction data, to initial soil bin pressure prediction model carries out migration learning;S6, using final soil bin pressure prediction model to the soil bin pressure of target domain shield machine is predicted in real time, determine whether the predicted soil bin pressure is reasonable, to feedback adjustment driving parameter.The application uses migration learning neural network model, improves the accuracy and prediction speed of target domain shield pressure prediction, and considers the influence on economic benefit under different driving parameter value conditions, to maximize economic benefit optimization shield driving parameter.

Description

Technical Field

[0001] This invention relates to the field of tunnel shield construction technology, specifically to a method for real-time prediction and adjustment of soil pressure in a tunnel boring machine based on transfer learning. Background Technology

[0002] Shield tunneling technology boasts high safety, high construction efficiency, and better economic benefits compared to mining methods, making it widely used in urban rail transit construction. A crucial guarantee for safe construction is maintaining the stability of the excavation face during shield tunneling. The shield cuts the soil in front of it with its cutterhead, and the excavated soil flows smoothly into the shield's soil chamber through openings on the cutterhead. The excavated soil in the chamber creates a certain pressure that balances with the pressure on the tunnel face, thus maintaining stability. When the actual monitored soil chamber pressure is lower than the set value, ground subsidence can occur. In this case, the screw conveyor speed can be reduced, the excavated soil discharge rate can be decreased, and the soil chamber pressure can be increased. Conversely, when the actual monitored soil chamber pressure is higher than the set value, ground uplift can occur in front of the shield. In this case, the screw conveyor speed can be increased, the excavated soil discharge rate can be increased, and the soil chamber pressure can be decreased. Therefore, setting an appropriate soil chamber pressure value is crucial during shield tunneling.

[0003] Currently, tunnel boring machines (TBMs) mainly rely on experience to determine the soil pressure value based on the conditions of the working strata. The pressure inside the soil chamber is monitored in real time by pressure sensors. However, it is impossible to reasonably predict changes in the soil pressure and therefore cannot predict and adjust the tunneling parameters in advance. Adjusting the tunneling parameters based on real-time monitoring data of the soil pressure inside the soil chamber has a certain lag, which makes it difficult to ensure the safety of TBM construction, meet the requirements of surface settlement control, and also hinders the timely adjustment and selection of appropriate tunneling parameters to reduce tunneling costs and meet economic requirements. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention proposes a real-time prediction and adjustment method for soil pressure in tunnel boring machines (TBMs) based on transfer learning. By rationally constructing a soil pressure prediction model, the method can efficiently and accurately predict the soil pressure value within the TBM area and adjust the TBM tunneling parameters based on the predicted soil pressure value, thereby greatly improving the safety and economic efficiency of TBM construction.

[0005] To achieve the above objectives, this invention provides a method for real-time prediction and adjustment of soil pressure in a tunnel boring machine (TBM) based on transfer learning, comprising the following steps:

[0006] S1. Determine if there is a pre-trained source domain task prediction model: If yes, use it as the initial earth pressure prediction model and proceed to step S5; if no, continue to the next step; wherein, the number of input parameters and the number of output parameters of the source domain task prediction model are the same as the number of input parameters and the number of output parameters used for target domain earth pressure prediction.

[0007] S2. Collect shield tunneling parameters, stratum type, stratum parameters, shield burial depth, stratum settlement monitoring values, and corresponding soil chamber pressure values. Preprocess the collected dataset using outlier detection and normalization to construct a training sample set. The shield tunneling parameters include: soil chamber pressure, total thrust, cutterhead torque, penetration depth, propulsion speed, cutterhead rotation speed, screw conveyor rotation speed, bentonite usage monitoring values, foam concentrate usage monitoring values, industrial water consumption monitoring values, and electricity consumption monitoring values.

[0008] S3. Construct an economic benefit evaluation index W to evaluate the direct economic losses caused by different tunneling parameter values:

[0009] W=α1w1+α2w2+α3w3+α4w4;

[0010] Among them, w1 is the bentonite usage monitoring value, α1 is the bentonite unit price, w2 is the foam raw material usage monitoring value, α2 is the foam raw material unit price, w3 is the industrial water consumption monitoring value, α3 is the industrial water unit price, w4 is the electricity consumption monitoring value, and α4 is the electricity unit price.

[0011] S4. Using the shield tunneling parameters, stratum type, stratum parameters, shield burial depth, and stratum settlement monitoring values ​​in the training sample set in step S2 as inputs and the soil pressure value in the training sample as output, train the pre-built neural network prediction model to obtain the initial soil pressure prediction model.

[0012] S5. Using the on-site construction data of the tunnel boring machine, transfer learning is performed on the initial soil chamber pressure prediction model. The weights and biases in the initial soil chamber pressure prediction model are corrected to obtain the final soil chamber pressure prediction model suitable for the site.

[0013] S6. Using the final earth chamber pressure prediction model, based on real-time shield tunneling parameters, geological type, geological parameters, shield burial depth, and ground settlement monitoring values, the earth chamber pressure of the shield machine in the target area is predicted in real time, and it is determined whether the predicted earth chamber pressure is between the upper and lower limits, so as to provide feedback for adjusting the tunneling parameters:

[0014] S61. When the predicted earth pressure value of the shield tunnel is between the upper and lower limits, the tunneling should continue with the corresponding tunneling parameters.

[0015] S62. When the predicted earth pressure value of the shield tunnel is not between the upper and lower limits, the range of tunneling parameters is predicted in reverse by the final earth pressure prediction model based on the upper and lower limits of the earth pressure. The economic benefit index W is calculated within the range of the tunneling parameters. The tunneling parameters corresponding to the minimum value of W are used as the appropriate tunneling parameters for tunneling.

[0016] Preferably, the stratigraphic types in the stratigraphic step S2 are classified as fractured zones, sand and gravel strata, and clay strata.

[0017] Preferably, the formation parameters in the formation step S2 include soil density, cohesion, internal friction angle, void ratio, and elastic modulus.

[0018] Preferably, the outlier detection processing of the collected dataset in step S2 uses the 3σ method, and the processing steps are as follows:

[0019] For each type of data in the collected dataset, including various tunneling parameters, geological parameters, shield burial depth, ground settlement monitoring values, and soil pressure values, their respective mean values ​​were calculated. and standard value σ i , where i is used to represent different types of data;

[0020] When the data in the dataset or When, then determine the data x i,j If the data is abnormal, then the data x is considered abnormal. i,j This represents normal data; the subscript j is used to distinguish different data of the same type.

[0021] Preferably, different processing methods are used for the abnormal data based on the amount of data collected:

[0022] When the amount of collected data exceeds 300 sets, the abnormal data and the corresponding remaining data will be removed.

[0023] When the amount of collected data is ≤300 sets, abnormal data shall be handled in the following manner:

[0024]

[0025] In the formula, x i,j and x' i,j These are the original data and the outlier-processed data corresponding to the i-th type in the j-th group of the dataset, respectively; the outlier-processed data x' i,j Used as normal data.

[0026] Preferably, in step S4, the pre-built neural network uses a grid search algorithm to optimize the number of neurons in each layer and the number of network layers to obtain the optimal parameters of the pre-built neural network prediction model.

[0027] Preferably, the calculation formulas for the upper and lower limits of the soil pressure in step S6 are as follows:

[0028] Upper limit of earth pressure: P fu =P1 + P2 + P3;

[0029] Lower limit of earth pressure: P f1 =P1+P2'+P3;

[0030] Wherein, P1: River water pressure (kPa);

[0031] P2: Earth pressure at rest (kPa);

[0032] P'2: Active earth pressure (kPa);

[0033] P3: Preparatory pressure, set to 20 kPa

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. This invention provides a real-time prediction and adjustment method for tunnel boring machine (TBM) soil pressure based on transfer learning. Through transfer learning, the weight parameters and bias parameters of other prediction network models (source domain task prediction models), or the weight parameters and bias parameters of soil pressure prediction models (source domain task prediction models) of other TBM projects, are applied to the prediction model (target domain task prediction model) of soil pressure in the target domain. This eliminates the need for initialization of the on-site TBM soil pressure prediction model, allows for early acquisition of changes in TBM soil pressure, and enables prediction of soil pressure values. This improves the prediction accuracy and timeliness of the on-site TBM soil pressure prediction model, enabling accurate and rapid prediction of soil pressure, ensuring tunnel construction safety, and ensuring that the surface settlement value in the shield tunnel area meets the requirements. It has good prediction timeliness and economy.

[0036] 2. This invention utilizes a transfer learning neural network model, which can transfer learn the relevant weight parameters of other source domain task prediction models when the amount of existing tunneling parameter data is small. This reduces the dependence on the amount of existing tunneling parameter data, significantly reduces the time required to establish a soil pressure prediction model, and improves the accuracy and speed of predicting shield pressure in the target domain.

[0037] 3. This invention can predict the pressure in the earth chamber in advance by taking tunneling parameters and geological parameters, understand the fluctuation of the pressure in the earth chamber, plan and adjust appropriate tunneling parameters in advance, and take corresponding safety measures to ensure the safe and smooth tunneling of the shield.

[0038] 4. This invention combines the economic costs generated during shield tunneling, accurately predicts the pressure of the earth chamber, and considers the impact of different tunneling parameter values ​​on economic benefits. It optimizes the tunneling parameters of the shield to maximize economic benefits, thus solving the problem that existing earth chamber prediction methods do not consider the actual construction costs of the project. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of the invention of a method for real-time prediction and adjustment of soil pressure in tunnel boring machines based on transfer learning;

[0041] Figure 2 This is a structural diagram of the neural network prediction model of the present invention;

[0042] Figure 3 It is a traditional machine learning method;

[0043] Figure 4 This invention relates to the transfer learning method for the earth pressure prediction model of the earth chamber. Detailed Implementation

[0044] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0045] The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the invention. Any modifications or substitutions made to the methods, steps, or conditions of the present invention without departing from the spirit and essence of the invention are within the scope of the invention.

[0046] like Figure 1 As shown in the figure, this embodiment provides a method for real-time prediction and adjustment of soil pressure in a tunnel boring machine based on transfer learning, including the following steps:

[0047] 1. Determine if there is a pre-trained source domain task prediction model: If yes, use it as the initial earth pressure prediction model and proceed to step S5; otherwise, continue to the next step; wherein, the number of input parameters and the number of output parameters of the source domain task prediction model are the same as the number of input parameters and the number of output parameters used for target domain earth pressure prediction.

[0048] 2. Collect shield tunneling parameters, geological strata types, geological parameters, shield burial depth, ground settlement monitoring values, and corresponding soil chamber pressure values. Preprocess the collected dataset using outlier detection and normalization to construct a training sample set. Shield tunneling parameters include: soil chamber pressure, total thrust, cutterhead torque, penetration depth, propulsion speed, cutterhead rotation speed, screw conveyor rotation speed, bentonite usage monitoring values, foam concentrate usage monitoring values, industrial water consumption monitoring values, and electricity consumption monitoring values. Geological strata types are classified as fractured zones, gravel strata, and clay strata. Geological parameters include soil density, cohesion, internal friction angle, void ratio, and elastic modulus.

[0049] The 3σ method was used to handle outlier detection in the collected dataset.

[0050] For each type of data in the collected dataset, including various tunneling parameters, geological parameters, shield burial depth, ground settlement monitoring values, and soil pressure values, their respective mean values ​​were calculated. and standard value σ i , where i takes values ​​of 1, 2, 3, 4, and 5, representing tunneling parameters, stratum parameters, shield burial depth, stratum settlement monitoring value, and soil chamber pressure value, respectively;

[0051] When the data in the dataset or When, then determine the data x i,j If the data is abnormal, then the data x is considered abnormal. i,j This represents normal data; the subscript j is used to distinguish different data of the same type.

[0052] Then, different processing methods are used for abnormal data based on the amount of data collected:

[0053] When the number of collected data sets is greater than 300, the sample data is sufficient, so the abnormal data and the corresponding remaining data are removed.

[0054] When the number of collected data sets is ≤300, the sample size is too small. Abnormal data should be handled as follows:

[0055]

[0056] In the formula, x i,j and x' i,jThese represent the original data and the outlier-processed data corresponding to the i-th type in the j-th group of the dataset, respectively. Normalizing the outlier-processed data allows us to construct the training samples, thus obtaining the training sample set. The outlier-processed data x' is the result of this outlier-processing. i,j Used as normal data to ensure the amount of data in the training sample set.

[0057] 3. Construct an economic benefit evaluation index W to evaluate the direct economic losses caused by different tunneling parameter values:

[0058] W=α1w1+α2w2+α3w3+α4w4;

[0059] Where w1 is the monitored value of bentonite usage, α1 is the unit price of bentonite, w2 is the monitored value of foam raw material usage, α2 is the unit price of foam raw material, w3 is the monitored value of industrial water consumption, α3 is the unit price of industrial water, w4 is the monitored value of electricity consumption, and α4 is the unit price of electricity consumption.

[0060] 4. Using the shield tunneling parameters, stratum type, stratum parameters, shield burial depth, and stratum settlement monitoring values ​​from the training sample set in step 2 as inputs, and the soil pressure value from the training samples as output, train the pre-constructed neural network prediction model (the structure of the constructed neural network prediction model is as follows). Figure 2 As shown in the figure, the initial earth chamber pressure prediction model is obtained.

[0061] A grid search algorithm is used to optimize the number of neurons in each layer and the number of network layers in the neural network prediction model to obtain the optimal parameters of the pre-constructed neural network prediction model. This step is then used to train and obtain the initial earth pressure prediction model.

[0062] The loss function used in this embodiment to train the neural network prediction model is:

[0063]

[0064] In the formula, Let P be the actual soil pressure value in the j-th training sample. j N is the earth pressure value predicted by the neural network prediction model based on the shield tunneling parameters, stratum type, stratum parameters, and shield burial depth of the j-th training sample. N is the number of samples used to train the neural network prediction model.

[0065] In addition, this embodiment uses K-fold cross-validation to obtain the model with the minimum loss function value as the optimal prediction model, thereby obtaining the optimal initial soil pressure prediction model.

[0066] 5. Using on-site construction data from the tunnel boring machine, transfer learning is performed on the initial soil chamber pressure prediction model. The transfer learning method is illustrated below. Figure 4 As shown, a traditional machine learning training method is illustrated as follows: Figure 3 As shown, by modifying the weights and biases in the initial earth chamber pressure prediction model through transfer learning, a final earth chamber pressure prediction model suitable for the field is obtained, which can improve the efficiency and accuracy of model training.

[0067] Transfer learning refers to the process of using knowledge from Ds and Ts to improve the prediction performance of the learning function for the target task, given a source domain Ds and a source task Ts, a target domain Dt and a target task Tt, even when Ds ≠ Dt or Ts ≠ Tt.

[0068] For example, when using source domain prediction models to predict cutterhead torque and total thrust based on tunneling parameters, and when the number of nodes in each model is the same, this model can be directly transferred to the prediction of tunnel boring machine soil pressure in the target domain of this invention. In other prediction models, when the input variables are different but the output variables are the same, the output layer parameters of the prediction model can be transferred to train the parameters of other layers. When the input layer of the prediction model is the same but the output layer is different, the input layer parameters can be transferred to train the parameters of the output layer model, thereby reducing the number of model parameters to train, reducing model training time, and improving the accuracy of model predictions.

[0069] 6. Using the final earth chamber pressure prediction model, based on real-time shield tunneling parameters, geological type, geological parameters, shield burial depth, and ground settlement monitoring values, the earth chamber pressure of the shield machine in the target area is predicted in real time. It is then determined whether the predicted earth chamber pressure is between the upper and lower limits to provide feedback for adjusting tunneling parameters.

[0070] If the predicted earth pressure value of the shield tunnel is between the upper and lower limits, then the tunneling should continue with the corresponding tunneling parameters.

[0071] When the predicted earth pressure value of the tunnel boring machine is not between the upper and lower limits, the range of tunneling parameters is predicted in reverse by the final earth pressure prediction model based on the upper and lower limits of the earth pressure. Within the range of tunneling parameters, the economic benefit index W is calculated. When W is at its minimum, the corresponding tunneling parameters are used as the appropriate tunneling parameters for tunneling.

[0072] The formulas for calculating the upper and lower limits of the earth pressure are as follows:

[0073] (1) Calculation of the upper limit value

[0074] Warehouse pressure limit: P fu =P1+P2+P3=γ W ×h+∑K0[(γ-γw )×d]+20

[0075] P fu : Upper limit of water pressure at the incision site (kPa);

[0076] P1: River water pressure (kPa);

[0077] P2: Earth pressure at rest (kPa);

[0078] P3: Preparatory pressure, typically 20 kPa;

[0079] d: Thickness of each soil layer

[0080] γ w : Specific weight of water (kN / m3);

[0081] h_water: The depth of the river water should be determined based on the tide table to determine the actual water depth (m) above the shield tunnel cut.

[0082] K0: Coefficient of earth pressure at rest;

[0083] γ: Soil bulk density (kN / m3);

[0084] H(h): Tunnel depth (calculated to the tunnel center) (m);

[0085] (2) Calculation of lower limit value

[0086] Lower limit of storage pressure: P f1 =P1+P2'+P3

[0087]

[0088] P f1 Lower limit of incision water pressure (kPa);

[0089] P'2: Active earth pressure (kPa);

[0090] K a Active earth pressure coefficient;

[0091] C u : Soil cohesion (kPa).

[0092] A subway phase II project's right line uses shield tunneling. Based on the methods described above, the earth-rock chamber pressure during the shield tunneling process was predicted, and the tunneling parameters were adjusted accordingly. The shield tunneling area is located in moderately weathered gravelly coarse sand strata, a type of sand and gravel stratum. The shield burial depth is 21.3 meters. When the shield is at ring 525, the tunneling parameters and predicted earth-rock chamber pressure are shown in Table 1:

[0093] Table 1. Tunneling parameters and predicted soil pressure of the shield tunnel.

[0094]

[0095] Based on the geological conditions, the theoretical upper limit of the soil pressure in this ring is calculated to be 2.5 Bar, and the lower limit is 2.7 Bar. The predicted soil pressure is higher than the theoretically calculated upper limit. The economic benefit W is calculated to be 1500 yuan.

[0096] Based on the earth pressure range of [2.5 Bar-2.7 Bar], the tunneling parameters were reverse-predicted. Based on the predicted tunneling parameter range and combined with the calculation of economic efficiency W, the final tunneling parameters were adjusted as shown in Table 2.

[0097] Table 2. Adjusted tunneling parameters of the shield tunnel and predicted earth pressure.

[0098]

[0099] As shown in Table 2, the tunneling parameters are relatively stable at this time, and the predicted soil pressure is consistent with the actual theoretically calculated soil pressure value. Moreover, the economic benefit W reaches 1210 yuan, and the economic loss is minimized. Under the premise of ensuring the safety of the tunnel boring machine, the cost is saved and the economic benefits are high.

[0100] The above-described embodiments are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the technical scope of the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for real-time prediction and adjustment of soil pressure in a tunnel boring machine (TBM) based on transfer learning, characterized in that, Includes the following steps: S1. Determine if there is a pre-trained source domain task prediction model: If yes, use it as the initial earth pressure prediction model and proceed to step S5; if no, continue to the next step; wherein, the number of input parameters and the number of output parameters of the source domain task prediction model are the same as the number of input parameters and the number of output parameters used for target domain earth pressure prediction. S2. Collect shield tunneling parameters, stratum type, stratum parameters, shield burial depth, stratum settlement monitoring values, and corresponding soil chamber pressure values. Preprocess the collected dataset using outlier detection and normalization to construct a training sample set. The shield tunneling parameters include: soil chamber pressure, total thrust, cutterhead torque, penetration depth, propulsion speed, cutterhead rotation speed, screw conveyor rotation speed, bentonite usage monitoring values, foam concentrate usage monitoring values, industrial water consumption monitoring values, and electricity consumption monitoring values. S3. Construct an economic benefit evaluation index W to evaluate the direct economic losses caused by different tunneling parameter values: ； in, This represents the monitoring value for bentonite usage. This refers to the unit price of bentonite. This represents the monitoring value for the amount of foam concentrate used. This refers to the unit price of the foam concentrate. This is the monitoring value for industrial water consumption. This is the unit price for industrial water. This is the electricity consumption monitoring value. This refers to the unit price of electricity. S4. Using the shield tunneling parameters, stratum type, stratum parameters, shield burial depth, and stratum settlement monitoring values ​​in the training sample set in step S2 as inputs and the soil pressure value in the training sample as output, train the pre-built neural network prediction model to obtain the initial soil pressure prediction model. S5. Using the on-site construction data of the tunnel boring machine, transfer learning is performed on the initial soil chamber pressure prediction model. The weights and biases in the initial soil chamber pressure prediction model are corrected to obtain the final soil chamber pressure prediction model suitable for the site. S6. Using the final earth chamber pressure prediction model, based on real-time shield tunneling parameters, geological type, geological parameters, shield burial depth, and ground settlement monitoring values, the earth chamber pressure of the shield machine in the target area is predicted in real time, and it is determined whether the predicted earth chamber pressure is between the upper and lower limits, so as to provide feedback for adjusting the tunneling parameters: S61. When the predicted earth pressure value of the shield tunnel is between the upper and lower limits, the tunneling should continue with the corresponding tunneling parameters. S62. When the predicted earth pressure value of the shield tunnel is not between the upper and lower limits, the range of tunneling parameters is predicted in reverse by the final earth pressure prediction model based on the upper and lower limits of the earth pressure. The economic benefit index W is calculated within the range of the tunneling parameters. The tunneling parameters corresponding to the minimum value of W are used as the appropriate tunneling parameters for tunneling.

2. The method for real-time prediction and adjustment of tunnel boring machine soil chamber pressure based on transfer learning according to claim 1, characterized in that, The geological types in step S2 are classified as fractured zones, sand and gravel strata, and clay strata.

3. The method for real-time prediction and adjustment of tunnel boring machine soil chamber pressure based on transfer learning according to claim 1, characterized in that, The formation parameters in step S2 include soil density, cohesion, internal friction angle, void ratio, and elastic modulus.

4. The method for real-time prediction and adjustment of tunnel boring machine soil chamber pressure based on transfer learning according to claim 1, characterized in that, The outlier detection processing of the collected dataset in step S2 adopts... The method and processing steps are as follows: For each type of data in the collected dataset, including various tunneling parameters, geological parameters, shield burial depth, ground settlement monitoring values, and soil pressure values, their respective mean values ​​were calculated. and standard values ,in Used to represent different types of data; When the data in the dataset or When, then determine the data. If the data is abnormal, otherwise the data is considered abnormal. Normal data; subscript Used to distinguish different data of the same type.

5. The method for real-time prediction and adjustment of tunnel boring machine soil chamber pressure based on transfer learning according to claim 4, characterized in that, Different processing methods are applied to the abnormal data based on the amount of data collected: When the amount of collected data exceeds 300 sets, the abnormal data and the corresponding remaining data will be removed. When the amount of collected data is ≤300 sets, abnormal data shall be handled in the following manner: ， In the formula, and The first in the dataset The first set of data The original data and outlier processing data corresponding to each type; outlier processing data Used as normal data.

6. The method for real-time prediction and adjustment of tunnel boring machine soil chamber pressure based on transfer learning according to claim 1, characterized in that, In step S4, the pre-built neural network uses a grid search algorithm to optimize the number of neurons in each layer and the number of network layers to obtain the optimal parameters of the pre-built neural network prediction model.

7. The method for real-time prediction and adjustment of tunnel boring machine soil chamber pressure based on transfer learning according to claim 1, characterized in that, The formulas for calculating the upper and lower limits of the soil pressure in step S6 are as follows: Upper limit of earth pressure: , Lower limit of earth pressure: , in, River water pressure (kPa); Earth pressure at rest (kPa); Active earth pressure (kPa); : Prepare pressure, take 20 kPa.

Images