Complex geological slurry shield cutter head and cutter type selection method based on deep learning

Abstract

The invention discloses a complex geological slurry shield cutterhead and cutter type selection method based on deep learning. The method comprises the steps of extraction of existing case information, extraction of current project information, establishment of a deep neural network model, training of the deep neural network model, prediction of cutterhead and cutter type selection and dynamic adjustment of a type selection scheme. The association between the geological conditions around the shield tunnel and the slurry shield cutterhead cutter type selection is excavated through the deep neural network model technology, and the internal influence process does not need to be considered, so that the complexity of the problem is remarkably reduced; therefore, targeted, efficient and accurate arrangement can be carried out on the slurry shield cutter head and the cutter according to complex geological conditions and local conditions; the cutterhead aperture ratio and the cutter loss rate are used as evaluation parts of the cost function of the deep neural network model, so that cutterhead characteristics and cutter use economy are fully considered when the deep neural network model is used for predicting slurry shield cutterhead cutter type selection.

Description

Technical Field

[0001] This invention belongs to the field of neural network model technology and relates to a method for selecting cutterheads for complex geological slurry shield tunnels based on deep learning. Background Technology

[0002] Deep learning algorithms, as a machine learning methodology based on artificial neural networks, automatically mine the inherent features of data and explore the optimal mathematical model for solving specific tasks through a multi-layered neural network architecture and training on large-scale datasets. Deep neural network model technology is one of the core technologies in the field of deep learning. Its characteristic lies in the presence of multiple hidden layers between the input and output layers, enabling the progressive extraction and abstract representation of data features layer by layer. With the improvement of computing power, deep neural network model technology is widely used in various infrastructure construction scenarios, especially in the selection of cutterheads for slurry-circulating tunnel boring machines (TBMs), which heavily relies on the experience of technical personnel.

[0003] The cutterhead and cutters of a slurry-circulating shield tunneling machine are used for breaking and excavating strata. The cutterhead is a key component located at the front of the tunneling machine, and the cutters are the cutting tools mounted on the cutterhead. Common cutterhead structures include panel type and spoke type, and common cutter types include hobbing cutters, cutting cutters, scrapers, and toothed cutters. During the arrival of the slurry-circulating shield tunneling machine or cutterhead replacement operations, the selection of cutterhead and cutters must be carried out in advance. This requires full consideration of the combined effects of geological conditions, construction requirements, cutter performance, and cutterhead structure.

[0004] Geological conditions are a key factor in determining the selection of cutterheads for slurry shield tunneling machines. Different geological formations impose varying requirements on cutter performance, such as wear resistance, impact toughness, and cutting efficiency. Therefore, it is necessary to study the correlation between key parameters of slurry shield cutters, including dimensions, materials, and arrangement of cutterheads, cutterhead materials, and cutterhead opening ratio, and geological conditions. Furthermore, considering engineering geological conditions, formation strength, and the presence of isolated boulders, research should be conducted on cutter selection based on formation wear resistance to reduce cutter wear rate, decrease cutter replacement frequency, and improve tunneling efficiency. However, in engineering practice, cutter selection for slurry shield tunneling machines primarily relies on the experience of technical personnel and reference to similar engineering cases. This makes it difficult to tailor and precisely arrange cutterheads according to complex geological conditions, thus compromising the efficiency and rationality of cutter selection for complex geological slurry shield tunneling machines.

[0005] Furthermore, in actual tunnel boring machine (TBM) construction, geological survey data often fails to fully and accurately reflect the complexity and variability of geological conditions, leading to discrepancies between the survey results and the actual geological situation. The selection of cutterhead tools for slurry TBMs based on geological survey data may not be fully adapted to the geological conditions encountered during actual tunneling. For example, when undetected hard rock interlayers or weak strata exist in the actual formation, the originally selected cutter may not be able to effectively break the strata or withstand significant impact forces, resulting in premature tool wear, damage, or low tunneling efficiency. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes a deep learning-based method for selecting cutterheads for slurry shield tunneling machines in complex geological conditions. This method is applicable to the selection of cutterheads for slurry shield tunneling machines under complex geological conditions, and helps to reduce cutter wear rate, reduce the number of cutter replacements, and improve tunneling efficiency, thereby achieving intelligent, efficient, and precise selection of cutterheads for slurry shield tunneling machines in complex geological conditions.

[0007] This application discloses a method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning. The method includes the following steps:

[0008] S100 Extraction of existing case information: Collect existing shield tunneling project case information, extract the selection of cutterhead and its influencing parameters of slurry shield tunneling machines from the existing shield tunneling project case information, and obtain the first selection parameter and the first influencing parameter;

[0009] S200 Extraction of current project information: Extract the characteristic parameters that affect the selection of cutterhead tools for slurry shield tunneling from the current survey and design results of the shield tunneling project to obtain the second influencing parameter;

[0010] S300. Establishment of a deep neural network model: Establish an initial deep neural network model, including an input layer, hidden layers, and an output layer; the input layer is the first layer of the initial deep neural network model, used to receive input data; the output layer is the last layer of the initial deep neural network model, used to generate prediction results; the hidden layer is the layer located between the input layer and the output layer, used for data processing and feature extraction;

[0011] S400. Training of the deep neural network model: Using the first influencing parameter as the input value and the first selection parameter as the output value, the initial deep neural network model is trained to obtain the trained deep neural network model.

[0012] S500, Prediction of cutterhead and tool selection: The second influencing parameter is input into the trained deep neural network model to predict the cutterhead and tool selection of the slurry shield tunneling machine at the start of the current shield tunneling project, to obtain the second selection parameter, and to formulate the first slurry shield tunneling machine cutterhead and tool selection scheme at the start of the shield tunneling project based on the second selection parameter.

[0013] S600. Dynamic adjustment of the selection scheme: When the cutter wear rate reaches the threshold, the cutter replacement operation is initiated. The characteristic parameters affecting the cutter selection of the slurry shield cutterhead are obtained at this time, and the third influencing parameter is obtained. The third influencing parameter is input into the trained deep neural network model to predict the cutter selection of the slurry shield cutterhead during the current cutter replacement operation of the shield project, and the third selection parameter is obtained. Based on the third selection parameter, the second slurry shield cutterhead cutter selection scheme during the cutter replacement operation of the shield project is formulated. The above operation of step S600 is repeated until the shield project is completed.

[0014] Preferably, the first selection parameter, the second selection parameter, and the third selection parameter all include tool size and structure, tool material, tool life, tool arrangement position, and tool head arrangement.

[0015] Preferably, the first influencing parameter, the second influencing parameter, and the third influencing parameter all include geological parameters and shield tunneling parameters; the geological parameters include, but are not limited to, stratum type, void ratio, compressive strength, water content, natural unit weight, static lateral pressure coefficient, saturation, liquid limit, plastic limit, characteristic value of stratum bearing capacity, internal friction angle, cohesion, horizontal permeability coefficient, and vertical permeability coefficient; the shield tunneling parameters include, but are not limited to, rated torque, escape torque, maximum working pressure, maximum tunneling speed, maximum thrust, and cutterhead drive power.

[0016] Preferably, when the first influencing parameter is input to the initial deep neural network model, and when both the second influencing parameter and the third influencing parameter are input to the trained deep neural network model, data normalization is performed first, and the normalized data x' satisfies the following expression:

[0017]

[0018] Where x is the data before normalization, x max x is the maximum value that this data can reach. min This is the minimum value that the data can reach.

[0019] Preferably, the evaluation part of the cost function of the initial deep neural network model includes: the difference between the model output and the first selection parameter; the difference between the tool head opening ratio calculated based on the model output and the tool head opening ratio calculated based on the first selection parameter; and the difference between the tool wear rate calculated based on the model output and the tool wear rate calculated based on the first selection parameter; the value of each evaluation part in the cost function is the weight corresponding to each evaluation part; the tool head opening ratio P satisfies the following expression:

[0020]

[0021] Among them, A r A represents the actual working area of ​​the cutter head. a This represents the total area of ​​the cutter head;

[0022] The tool wear rate L satisfies the following expression:

[0023]

[0024] Where, N r N represents the number of worn-out tools. a This represents the total number of cutting tools.

[0025] Preferably, the trained deep network model is fine-tuned using a stochastic gradient descent algorithm or a regularization method.

[0026] Preferably, the monitoring of cutter wear rate during shield tunneling is achieved through a monitoring system, which includes an information acquisition module, an information transmission module, an information processing module, and an alarm module.

[0027] The information acquisition module includes an infrared thermal imager installed on the tunnel boring machine, which is used to indirectly reflect the wear status of the cutting tools by monitoring the temperature;

[0028] The information transmission module is used to establish data connections and transmissions between modules;

[0029] The information processing module includes a data preprocessing unit and an analysis and judgment unit; wherein, the data preprocessing unit is used to preprocess the raw data acquired by the information acquisition module to reduce data noise and redundancy; the analysis and judgment unit is used to calculate the tool wear rate and determine whether the tool wear rate exceeds the threshold.

[0030] The alarm module is used to control the alarm to sound an alarm based on the judgment result of the information processing module.

[0031] Compared with existing technologies, the beneficial effects of this invention are as follows: For the selection of cutterhead tools in slurry shield tunnels under complex geological conditions, a deep learning-based method for selecting cutterhead tools for complex geological slurry shield tunnels is proposed. This method includes extracting existing case information, extracting current project information, establishing a deep neural network model, training the deep neural network model, predicting cutterhead tool selection, and dynamically adjusting the selection scheme. By using deep neural network model technology to explore the correlation between the geological conditions surrounding the shield tunnel and the selection of cutterhead tools for slurry shield tunnels, the complexity of the problem is significantly reduced without considering the internal influencing processes. This allows for targeted, efficient, and precise arrangement of cutterhead tools based on complex geological conditions. By using the cutterhead opening ratio and tool wear rate as evaluation components of the deep neural network model's cost function, the deep neural network model fully considers the characteristics of the cutterhead and the economic efficiency of tool use when predicting cutterhead tool selection for slurry shield tunnels. The corresponding cutterhead tool selection for slurry shield tunnels is obtained through measured geological conditions during cutter replacement operations, ensuring that the selection scheme accurately reflects the complexity and variability of geological conditions. Attached Figure Description

[0032] Figure 1 This is a flowchart of the deep learning-based method for selecting cutterheads in complex geological slurry shield tunneling machines, as described in this invention.

[0033] Figure 2 This is a schematic diagram of the structure of a deep neural network model provided in an embodiment of the present invention;

[0034] Figure 3 This is a connection diagram of a monitoring system provided in an embodiment of the present invention;

[0035] Reference numerals: 1-Information acquisition module, 11-Infrared thermal imager, 2-Information transmission module, 3-Information processing module, 31-Data preprocessing unit, 32-Analysis and judgment unit, 4-Alarm module. Detailed Implementation

[0036] The following is in conjunction with the appendix Figure 1-3 The accompanying drawings and reference numerals provide a more detailed description of the embodiments of the present invention, enabling those skilled in the art to implement it after reading this specification. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0037] This application discloses, as follows: Figures 1 to 3 The deep learning-based method for selecting cutterheads in complex geological slurry shield tunneling machines includes the following steps:

[0038] S100 Extraction of existing case information: Collect existing shield tunneling project case information, extract the selection of cutterhead and its influencing parameters of slurry shield tunneling machines from the existing shield tunneling project case information, and obtain the first selection parameter and the first influencing parameter;

[0039] S200 Extraction of current project information: Extract the characteristic parameters that affect the selection of cutterhead tools for slurry shield tunneling from the current survey and design results of the shield tunneling project to obtain the second influencing parameter;

[0040] S300. Establishing the Deep Neural Network Model: Establish the initial deep neural network model, including the input layer (I1, I2, ... I... k ), hidden layer (h1) [1] h2 [1] ,…h n [3] ) and output layers (O1, O2, ... O j The input layer is the first layer of the initial deep neural network model, used to receive input data; the output layer is the last layer of the initial deep neural network model, used to generate prediction results; the hidden layer is the layer located between the input layer and the output layer, used for data processing and feature extraction.

[0041] S400. Training of the deep neural network model: Using the first influencing parameter as the input value and the first selection parameter as the output value, the initial deep neural network model is trained to obtain the trained deep neural network model.

[0042] S500, Prediction of cutterhead and tool selection: The second influencing parameter is input into the trained deep neural network model to predict the cutterhead and tool selection of the slurry shield tunneling machine at the start of the current shield tunneling project, to obtain the second selection parameter, and to formulate the first slurry shield tunneling machine cutterhead and tool selection scheme at the start of the shield tunneling project based on the second selection parameter.

[0043] S600. Dynamic Adjustment of the Selection Scheme: When the cutter wear rate reaches the threshold during construction monitoring, the cutter replacement operation is initiated. Characteristic parameters affecting the selection of cutterhead cutters for the slurry shield tunneling machine are obtained through geological surveys and indoor tests, resulting in a third influencing parameter. The third influencing parameter is input into the trained deep neural network model to predict the selection of cutterhead cutters for the slurry shield tunneling machine during the current cutter replacement operation, resulting in a third selection parameter. Based on the third selection parameter, a second slurry shield tunneling machine cutterhead selection scheme is formulated for the cutter replacement operation. The above operations of step S600 are repeated until the tunneling project is completed.

[0044] In practice, the first selection parameter, the second selection parameter, and the third selection parameter all include tool size and structure, tool material, tool life, tool arrangement position, and tool head arrangement.

[0045] In specific implementation, the first influencing parameter, the second influencing parameter, and the third influencing parameter all include geological parameters and shield tunneling parameters; the geological parameters include, but are not limited to, stratum type, void ratio, compressive strength, water content, natural unit weight, static lateral pressure coefficient, saturation, liquid limit, plastic limit, characteristic value of stratum bearing capacity, internal friction angle, cohesion, horizontal permeability coefficient, and vertical permeability coefficient; the shield tunneling parameters include, but are not limited to, rated torque, escape torque, maximum working pressure, maximum tunneling speed, maximum thrust, and cutterhead drive power.

[0046] In specific implementation, when the first influencing parameter is input to the initial deep neural network model, and when the second influencing parameter and the third influencing parameter are input to the trained deep neural network model, data normalization processing is first performed. The normalized data x' satisfies the following expression:

[0047]

[0048] Where x is the data before normalization, x max x is the maximum value that this data can reach. min This is the minimum value that the data can reach;

[0049] Under typical operating conditions, the saturation degree x before normalization is 0.8, and the maximum achievable saturation degree x is... max The minimum value of saturation that can be achieved is 1. min If the value is 0, then the normalized saturation x' is calculated according to equation (1) as follows:

[0050]

[0051] The normalized saturation x' can be input into the initial deep neural network model or the trained deep neural network model.

[0052] In specific implementation, the evaluation part of the cost function of the initial deep neural network model includes: the difference between the model output and the first selection parameter; the difference between the cutterhead opening ratio calculated based on the model output and the cutterhead opening ratio calculated based on the first selection parameter; and the difference between the tool wear rate calculated based on the model output and the tool wear rate calculated based on the first selection parameter. The value of each evaluation part in the cost function is its corresponding weight. By using the cutterhead opening ratio and tool wear rate as evaluation parts of the cost function of the deep neural network model, the deep neural network model fully considers the characteristics of the cutterhead and the economic efficiency of tool use when predicting the selection of cutterhead tools for slurry shield tunneling machines. The cutterhead opening ratio P satisfies the following expression:

[0053]

[0054] Among them, A r A represents the actual working area of ​​the cutter head. a This represents the total area of ​​the cutter head;

[0055] Under typical operating conditions, the actual working area A of the cutter head r It is 9.3m 2 The total area of ​​the cutter head is A a 31m 2 The cutterhead aperture ratio P is calculated according to equation (3) as follows:

[0056]

[0057] The tool wear rate L satisfies the following expression:

[0058]

[0059] Where, N r N represents the number of worn-out tools. a This represents the total number of cutting tools.

[0060] Under typical operating conditions, the number of worn tools, N r There are 3, and the total number of cutting tools is N. a The number of tools is 38, and the tool wear rate L is calculated according to equation (5) as follows:

[0061]

[0062] In practice, the trained deep network model is fine-tuned using a stochastic gradient descent algorithm or a regularization method.

[0063] In practice, the monitoring of cutter wear rate during shield tunneling is achieved through a monitoring system, which includes an information acquisition module 1, an information transmission module 2, an information processing module 3, and an alarm module 4.

[0064] The information acquisition module 1 includes an infrared thermal imager 11 installed on the tunnel boring machine, which is used to indirectly reflect the wear status of the cutting tools by monitoring the temperature;

[0065] The information transmission module 2 is used to establish data connections and transmissions between the modules;

[0066] The information processing module 3 includes a data preprocessing unit 31 and an analysis and judgment unit 32; wherein, the data preprocessing unit 31 is used to preprocess the raw data acquired by the information acquisition module 1 to reduce data noise and redundancy; the analysis and judgment unit 32 is used to calculate the tool wear rate and determine whether the tool wear rate exceeds the threshold.

[0067] The alarm module 4 is used to control the alarm to sound an alarm based on the judgment result of the information processing module 3.

[0068] Therefore, it is evident that by using deep neural network model technology to explore the correlation between the geological conditions surrounding the shield tunnel and the selection of cutterhead tools for slurry shield tunnels, the complexity of the problem is significantly reduced without considering the internal influencing processes. This allows for targeted, efficient, and precise arrangement of cutterhead tools for slurry shield tunnels based on complex geological conditions. By using the cutterhead opening ratio and tool wear rate as evaluation components of the cost function of the deep neural network model, the model fully considers the characteristics of the cutterhead and the economic efficiency of tool use when predicting the selection of cutterhead tools for slurry shield tunnels. By obtaining the corresponding cutterhead tool selection based on the measured geological conditions during the cutter replacement operation, the selection scheme can accurately reflect the complexity and variability of the geological conditions.

[0069] The above describes one or more embodiments of the present invention in a relatively specific and detailed manner, but it should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning, characterized in that, Includes the following steps: S100 Extraction of existing case information: Collect existing shield tunneling project case information, extract the selection of cutterhead and its influencing parameters of slurry shield tunneling machines from the existing shield tunneling project case information, and obtain the first selection parameter and the first influencing parameter; S200 Extraction of current project information: Extract the characteristic parameters that affect the selection of cutterhead tools for slurry shield tunneling from the current survey and design results of the shield tunneling project to obtain the second influencing parameter; S300. Establishment of a deep neural network model: Establish an initial deep neural network model, including an input layer, hidden layers, and an output layer; the input layer is the first layer of the initial deep neural network model, used to receive input data; the output layer is the last layer of the initial deep neural network model, used to generate prediction results; the hidden layer is the layer located between the input layer and the output layer, used for data processing and feature extraction; S400. Training of the deep neural network model: Using the first influencing parameter as the input value and the first selection parameter as the output value, the initial deep neural network model is trained to obtain the trained deep neural network model. S500, Prediction of cutterhead and tool selection: The second influencing parameter is input into the trained deep neural network model to predict the cutterhead and tool selection of the slurry shield tunneling machine at the start of the current shield tunneling project, to obtain the second selection parameter, and to formulate the first slurry shield tunneling machine cutterhead and tool selection scheme at the start of the shield tunneling project based on the second selection parameter. S600. Dynamic adjustment of the selection scheme: When the cutter wear rate reaches the threshold, the cutter replacement operation is initiated. The characteristic parameters affecting the cutter selection of the slurry shield cutterhead are obtained at this time, and the third influencing parameter is obtained. The third influencing parameter is input into the trained deep neural network model to predict the cutter selection of the slurry shield cutterhead during the current cutter replacement operation of the shield project, and the third selection parameter is obtained. Based on the third selection parameter, the second slurry shield cutterhead cutter selection scheme during the cutter replacement operation of the shield project is formulated. The above operation of step S600 is repeated until the shield project is completed.

2. The method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning as described in claim 1, characterized in that, The first selection parameter, the second selection parameter, and the third selection parameter all include tool size and structure, tool material, tool life, tool arrangement position, and tool head arrangement.

3. The method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning as described in claim 1, characterized in that, The first, second, and third influencing parameters all include geological parameters and tunnel boring machine (TBM) construction parameters. The geological parameters include, but are not limited to, stratum type, void ratio, compressive strength, water content, natural unit weight, static lateral pressure coefficient, saturation, liquid limit, plastic limit, characteristic value of stratum bearing capacity, internal friction angle, cohesion, horizontal permeability coefficient, and vertical permeability coefficient. The TBM construction parameters include, but are not limited to, rated torque, escape torque, maximum working pressure, maximum tunneling speed, maximum thrust, and cutterhead drive power.

4. The method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning according to claim 1, characterized in that, When the first influencing parameter is input to the initial deep neural network model, and when the second influencing parameter and the third influencing parameter are both input to the trained deep neural network model, data normalization is first performed. The normalized data x' satisfies the following expression: Where x is the data before normalization, x max x is the maximum value that this data can reach. min This is the minimum value that the data can reach.

5. The method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning according to claim 1, characterized in that, The evaluation portion of the cost function of the initial deep neural network model includes: the difference between the model output and the first selection parameter; the difference between the tool head opening ratio calculated based on the model output and the tool head opening ratio calculated based on the first selection parameter; and the difference between the tool wear rate calculated based on the model output and the tool wear rate calculated based on the first selection parameter. The values ​​of each evaluation portion in the cost function are the corresponding weights. The tool head opening ratio P satisfies the following expression: Among them, A r A represents the actual working area of ​​the cutter head. a This represents the total area of ​​the cutter head; The tool wear rate L satisfies the following expression: Where, N r N represents the number of worn-out tools. a This represents the total number of cutting tools.

6. The method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning according to claim 1, characterized in that, The trained deep network model is fine-tuned using a stochastic gradient descent algorithm or a regularization method.

7. The method for selecting cutterheads for complex geological slurry shield tunneling machines based on deep learning according to claim 1, characterized in that, The monitoring of cutter wear rate during shield tunneling is achieved through a monitoring system, which includes an information acquisition module, an information transmission module, an information processing module, and an alarm module. The information acquisition module includes an infrared thermal imager installed on the tunnel boring machine, which is used to indirectly reflect the wear status of the cutting tools by monitoring the temperature; The information transmission module is used to establish data connections and transmissions between modules; The information processing module includes a data preprocessing unit and an analysis and judgment unit; wherein, the data preprocessing unit is used to preprocess the raw data acquired by the information acquisition module to reduce data noise and redundancy; the analysis and judgment unit is used to calculate the tool wear rate and determine whether the tool wear rate exceeds the threshold. The alarm module is used to control the alarm to sound an alarm based on the judgment result of the information processing module.

Images