A cutter head mud cake detection method and device, electronic equipment and storage medium

Abstract

The application discloses a cutterhead mud cake detection method and device, electronic equipment and a storage medium, and belongs to the technical field of tunnel construction. The detection method comprises the following steps: reading historical tunneling data from a PLC controller and an industrial computer of a shield tunneling machine; training a parameter trend detection model by using mud cake related parameters in the historical tunneling data; the parameter trend detection model comprises a one-dimensional convolution layer, an LSTM layer and a full connection layer; obtaining mud cake related parameters of the shield tunneling machine in a target time period, inputting the mud cake related parameters in the target time period into the parameter trend detection model, and obtaining a change trend of the mud cake related parameters in the target time period; judging whether the change trend of the mud cake related parameters in the target time period conforms to cutterhead mud cake characteristics; if yes, it is determined that the shield tunneling machine is in a cutterhead mud cake state; and if no, it is determined that the shield tunneling machine is not in the cutterhead mud cake state. The application can improve the accuracy of detecting cutterhead mud cake.

Description

Technical Field

[0001] This application relates to the field of tunnel construction technology, and in particular to a method, apparatus, electronic device and storage medium for detecting mud cake on a cutterhead. Background Technology

[0002] As specialized engineering machinery used in tunnel and subway construction, the cutterhead is the core and critical component of a tunnel boring machine (TBM). During tunneling, if the operator lacks sufficient theoretical or practical experience and some performance parameters are set improperly, mud cake formation on the cutterhead can occur. Mud cake formation reduces the advance speed, lowers tunneling efficiency, and increases the cutterhead temperature. Increased cutterhead temperature can lead to localized burns, changes in material properties, and severe wear, resulting in a reduced service life.

[0003] In related technologies, a temperature acquisition module is used to collect the temperature of the cutterhead cutting end face, and the collected temperature is used to determine whether the tunnel boring machine is in a state of cutterhead mud cake formation. However, the environment in which the temperature acquisition module operates is relatively harsh, with strong vibrations, making it prone to damage. At the same time, it may be necessary to add water to the soil chamber, which may affect the accuracy of the real-time temperature measurement by the temperature sensor, making it impossible to accurately detect cutterhead mud cake formation.

[0004] Therefore, improving the accuracy of detecting sludge cake on the cutter head is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus, electronic device, and storage medium for detecting cutterhead cake, which can improve the accuracy of detecting cutterhead cake.

[0006] To address the aforementioned technical problems, this application provides a method for detecting clay cake buildup on a cutterhead, comprising:

[0007] Historical tunneling data is read from the PLC controller and industrial computer of the tunnel boring machine, and parameters related to mud cake are extracted from the historical tunneling data;

[0008] A parameter trend detection model is trained using the mud cake-related parameters in the historical tunneling data; wherein, the parameter trend detection model includes a one-dimensional convolutional layer, an LSTM layer, and a fully connected layer;

[0009] The relevant parameters of mud cake formation of the tunnel boring machine within the target time period are obtained, and the relevant parameters of mud cake formation within the target time period are input into the parameter trend detection model to obtain the changing trend of the relevant parameters of mud cake formation within the target time period.

[0010] Determine whether the changing trend of the mud cake-related parameters within the target time period conforms to the characteristics of cutterhead mud cake; if yes, determine that the tunnel boring machine is in the state of cutterhead mud cake; if no, determine that the tunnel boring machine is not in the state of cutterhead mud cake.

[0011] Optionally, parameters related to mud cake can be extracted from the historical tunneling data, including:

[0012] Extract mud cake-related parameters from the historical tunneling data that have a correlation greater than a preset value with the state of mud cake on the cutterhead; wherein, the mud cake-related parameters include propulsion speed, total propulsion force, and cutterhead torque.

[0013] Optionally, the parameters related to the sludge cake formation also include the cutter head rotation speed and penetration depth.

[0014] Optionally, a parameter trend detection model is trained using the mud cake-related parameters in the historical tunneling data, including:

[0015] Input the mud cake-related parameters from the historical tunneling data into the one-dimensional convolutional layer to obtain the local trend features of the mud cake-related parameters;

[0016] Input the mud cake-related parameters from the historical tunneling data into the LSTM layer to obtain the long-term trend characteristics of the mud cake-related parameters;

[0017] The parametric trend detection model is trained using the local trend features, the long-term trend features, and the objective function.

[0018] The local trend feature is used to describe the trend of change within a first time period, and the long-term trend feature is used to describe the trend of change within a second time period. The first time period is shorter than the second time period, and the trend of change includes the trend duration and slope.

[0019] Optionally, a parameter trend detection model is trained using the mud cake-related parameters in the historical tunneling data, including:

[0020] A training set is extracted from the historical tunneling data according to a preset ratio;

[0021] The parameters related to mud cake in the training set are normalized based on the maximum range value of the parameters related to mud cake.

[0022] The parameter trend detection model is trained using the normalized training set.

[0023] Optionally, after determining that the tunnel boring machine is in a state of mud cake formation on the cutterhead, the method further includes:

[0024] The tunnel boring machine is controlled to stop tunneling and an alarm message is output.

[0025] Optionally, if the parameters related to cake formation include propulsion speed, total propulsion force, and cutterhead torque, then determining whether the changing trend of the parameters related to cake formation within the target time period conforms to the characteristics of cutterhead cake formation includes:

[0026] Determine whether the changing trends of propulsion speed, total propulsion force, and cutterhead torque within the target time period conform to the characteristics of cutterhead cake formation.

[0027] The characteristics of the cutterhead forming mud cake are that the propulsion speed continuously decreases, the total propulsion force continuously increases, and the cutterhead torque continuously increases within the target time period.

[0028] This application also provides a device for detecting clay cake on a cutter head, comprising:

[0029] The parameter extraction module is used to read historical tunneling data from the PLC controller and industrial control computer of the tunnel boring machine, and extract mud cake-related parameters from the historical tunneling data;

[0030] The model training module is used to train a parameter trend detection model using the mud cake-related parameters in the historical tunneling data; wherein, the parameter trend detection model includes a one-dimensional convolutional layer, an LSTM layer, and a fully connected layer;

[0031] The trend detection module is used to obtain the mud cake-related parameters of the tunnel boring machine within a target time period, input the mud cake-related parameters within the target time period into the parameter trend detection model, and obtain the changing trend of the mud cake-related parameters within the target time period.

[0032] The judgment module is used to determine whether the changing trend of the mud cake-related parameters within the target time period conforms to the characteristics of cutterhead mud cake; if yes, the tunnel boring machine is determined to be in the state of cutterhead mud cake; if no, the tunnel boring machine is determined not to be in the state of cutterhead mud cake.

[0033] This application also provides a storage medium storing a computer program thereon, which, when executed, performs the steps of the above-described method for detecting sludge cake on the cutter head.

[0034] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor, when calling the computer program in the memory, implements the steps of the above-described method for detecting mud cake on the cutter head.

[0035] This application provides a method for detecting cutterhead mud cake, comprising: reading historical tunneling data from the PLC controller and industrial control computer of the tunnel boring machine (TBM), and extracting mud cake-related parameters from the historical tunneling data; training a parameter trend detection model using the mud cake-related parameters in the historical tunneling data; wherein the parameter trend detection model includes a one-dimensional convolutional layer, an LSTM layer, and a fully connected layer; obtaining the mud cake-related parameters of the TBM within a target time period, inputting the mud cake-related parameters within the target time period into the parameter trend detection model to obtain the changing trend of the mud cake-related parameters within the target time period; determining whether the changing trend of the mud cake-related parameters within the target time period conforms to the characteristics of cutterhead mud cake; if yes, determining that the TBM is in a cutterhead mud cake state; if no, determining that the TBM is not in a cutterhead mud cake state.

[0036] This application extracts mud cake-related parameters from historical tunneling data of tunnel boring machines (TBMs) and trains a parameter trend detection model using these parameters. This model outputs the changing trends of the mud cake-related parameters. After inputting the mud cake-related parameters for a target time period into the model, it outputs the changing trends within that period. Since these parameters are related to the mud cake state of the TBM cutterhead, the changing trends output by the model can be used to evaluate whether the TBM is in a mud cake state. Detecting the cutterhead mud cake state based on the changing trends of these parameters reduces interference from the working environment and improves the accuracy of mud cake detection. This application also provides a cutterhead mud cake detection device, a storage medium, and an electronic device, all with the aforementioned advantages, which will not be elaborated upon further here. Attached Figure Description

[0037] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating a method for detecting sludge cake on a cutter head, provided in an embodiment of this application;

[0039] Figure 2 A schematic diagram of a data transmission structure for a tunnel boring machine provided in an embodiment of this application;

[0040] Figure 3 A flowchart illustrating a parameter trend detection model trained and tested using tunnel boring machine excavation data, provided in an embodiment of this application;

[0041] Figure 4 A flowchart illustrating the model porting to a tunnel boring machine control system provided in this application embodiment;

[0042] Figure 5 This is a schematic diagram of the structure of a device for detecting mud cake on a cutter head, provided in an embodiment of this application. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] Please see below. Figure 1 , Figure 1 This is a flowchart illustrating a method for detecting mud cake on a cutter head, as provided in an embodiment of this application.

[0045] Specific steps may include:

[0046] S101: Read historical tunneling data from the PLC controller and industrial computer of the tunnel boring machine, and extract mud cake-related parameters from the historical tunneling data;

[0047] This embodiment can be applied to electronic devices connected to a tunnel boring machine. Please refer to [link to relevant documentation]. Figure 2 , Figure 2 This is a schematic diagram of a data transmission structure for a tunnel boring machine (TBM) according to an embodiment of this application. In the diagram, 201 represents a PLC controller, 202 represents an industrial control computer, and 203 represents a sensor. During the tunneling process, the TBM can use sensors to collect tunneling data from various parts, including parameters such as cutterhead speed, zone cylinder pressure, excavation chamber pressure, and working chamber pressure. The tunneling data can be transmitted between the sensors, PLC controller, and industrial control computer via Ethernet.

[0048] In this embodiment, the tunneling data of the tunnel boring machine can include various parameters. Some parameters are stored on the industrial control computer, while others are collected by sensors and can be read from the PLC controller. Prior to this step, there may be an operation to determine a historical time period, thereby setting the tunneling data stored in the PLC controller and industrial control computer within that historical time period as historical tunneling data.

[0049] Historical tunneling data can be used to identify known time periods when the cutterhead was in a mud-caked state. This historical data contains parameters related to cutterhead mud-caked conditions and parameters unrelated to it. This embodiment analyzes the correlation between each parameter in the historical tunneling data and the cutterhead mud-caked state, and identifies parameters with a correlation greater than a preset value as mud-caked-related parameters. As a feasible implementation, this step can use the Pearson correlation coefficient method to extract mud-caked-related parameters from the historical tunneling data that have a correlation greater than a preset value with the cutterhead mud-caked state. Specifically, these mud-caked-related parameters include propulsion speed, total propulsion force, and cutterhead torque. Furthermore, these mud-caked-related parameters may also include cutterhead rotational speed and penetration depth.

[0050] S102: Train a parameter trend detection model using the mud cake-related parameters in the historical tunneling data;

[0051] In this process, after determining the mud cake-related parameters in the historical tunneling data, and before training the parameter trend detection model using the mud cake-related parameters in the historical tunneling data, the mud cake-related parameters in the historical tunneling data can be normalized according to the maximum range value of the mud cake-related parameters, so as to use the normalized mud cake-related parameters to train the parameter trend detection model.

[0052] This step trains the parameter trend detection model using mud cake-related parameters from a historical time period. The parameter trend detection model may include one-dimensional convolutional layers, LSTM layers, and fully connected layers. The one-dimensional convolutional layers extract local trend features of the mud cake-related parameters, and the LSTM layers extract long-term trend features. After training the parameter trend detection model, it can calculate the changing trend of the input parameters. These changing trends include: no obvious trend, continuous increase, step increase, continuous decrease, and step decrease. The curves corresponding to continuous increase and continuous decrease do not have abrupt slopes, while the curves corresponding to step increase and step decrease have abrupt slope points.

[0053] S103: Obtain the mud cake-related parameters of the tunnel boring machine within the target time period, input the mud cake-related parameters within the target time period into the parameter trend detection model, and obtain the change trend of the mud cake-related parameters within the target time period.

[0054] Once the parameters highly correlated with the mud cake state are determined, if it is necessary to determine whether the tunnel boring machine (TBM) experiences cutterhead mud cake formation within a target time period, the relevant mud cake parameters for that time period can be directly obtained. There is no need to obtain parameters unrelated to cutterhead mud cake formation, thus improving detection efficiency. The target time period can be a continuous period without intervals. As a feasible implementation method, the target time period [t0-Δt, t0] can be determined based on the current time t0, where Δt is the duration of the target time period (e.g., 5 minutes, 10 minutes, etc.).

[0055] After obtaining the relevant parameters of mud cake within the target time period, the relevant parameters of mud cake collected within the target time period can be input into the parameter trend detection model so that the parameter trend detection model can output the change trend of the relevant parameters of mud cake within the target time period.

[0056] S104: Determine whether the changing trend of the parameters related to cake formation within the target time period conforms to the characteristics of cake formation on the cutterhead; if yes, proceed to S105; if no, proceed to S106.

[0057] S105: It is determined that the tunnel boring machine is in a state of mud cake formation on the cutterhead;

[0058] S106: It is determined that the tunnel boring machine is not in the state of mud cake formation on the cutterhead.

[0059] In this embodiment, a standard trend of change in parameters related to cutterhead mud cake formation can be preset when the tunnel boring machine (TBM) is in a mud cake formation state, i.e., the cutterhead mud cake characteristics. If the trend of change in parameters related to mud cake formation within a target time period matches the cutterhead mud cake characteristics, the TBM is determined to be in a cutterhead mud cake formation state; if the trend of change in parameters related to mud cake formation within the target time period does not match the cutterhead mud cake characteristics, the TBM is determined not to be in a cutterhead mud cake formation state. Furthermore, after determining that the TBM is in a cutterhead mud cake formation state, the TBM can be controlled to stop tunneling and an alarm message can be output to prevent cutterhead wear.

[0060] Taking the parameters related to mud cake formation (propulsion speed, total propulsion force, and cutterhead torque) as an example, the following methods can be used to determine whether the changing trends of these parameters within a target time period conform to the characteristics of cutterhead mud cake formation: determining whether the changing trends of propulsion speed, total propulsion force, and cutterhead torque within the target time period conform to the characteristics of cutterhead mud cake formation; wherein, the characteristics of cutterhead mud cake formation are a continuous decrease in propulsion speed, a continuous increase in total propulsion force, and a continuous increase in cutterhead torque within the target time period. The duration corresponding to the continuous increase and continuous decrease in the above-mentioned characteristics of cutterhead mud cake formation is a preset duration. If the preset duration is less than or equal to the duration corresponding to the target time period, and if within the target time period, the propulsion speed continuously decreases, the total propulsion force continuously increases, and the cutterhead torque continuously increases, then the tunnel boring machine is determined to be in a state of cutterhead mud cake formation; under other combinations of changing trends, the tunnel boring machine is determined not to be in a state of cutterhead mud cake formation.

[0061] This embodiment extracts mud cake-related parameters from historical tunneling data of the tunnel boring machine (TBM) and trains a parameter trend detection model using these parameters. This model outputs the changing trends of these parameters. After inputting the mud cake-related parameters for a target time period into the model, it outputs the changing trends within that period. Since these parameters are related to the mud cake state of the TBM cutterhead, the model's output trend can be used to evaluate whether the TBM is experiencing cutterhead mud cake formation. This process, detecting cutterhead mud cake formation based on the changing trends of these parameters, reduces interference from the working environment and improves the accuracy of cutterhead mud cake detection.

[0062] As for Figure 1 Further description of the corresponding embodiment: the parameter trend detection model includes two one-dimensional convolutional layers, two LSTM layers, and one fully connected layer. Therefore, the process of training the parameter trend detection model using mud cake-related parameters from historical tunneling data includes: inputting the mud cake-related parameters from the historical tunneling data into the one-dimensional convolutional layer to obtain local trend features of the mud cake-related parameters; inputting the mud cake-related parameters from the historical tunneling data into the LSTM layer to obtain long-term trend features of the mud cake-related parameters; and training the parameter trend detection model using the local trend features, the long-term trend features, and an objective function. The local trend features describe the trend within a first time period (e.g., 1 second), and the long-term trend features describe the trend within a second time period (e.g., 10 seconds). The first time period is shorter than the second time period, and the trend includes the trend duration and slope. The local trend features and the long-term trend features are feature vectors describing the parameter change trend.

[0063] As for Figure 1A further description of the corresponding embodiment is that the parameter trend detection model can be trained in the following way: extract a training set from the historical tunneling data according to a preset ratio; normalize the mud cake-related parameters in the training set according to the maximum range value of the mud cake-related parameters; and train the parameter trend detection model using the normalized training set.

[0064] Specifically, before training the parameter trend detection model using the mud cake-related parameters from the historical tunneling data, the mud cake-related parameters in the historical tunneling data can be divided into a training set and a test set according to a preset ratio. This preset ratio can be 4:1. Based on this, the parameter trend detection model can be trained using the mud cake-related parameters in the training set. After training the parameter trend detection model using the mud cake-related parameters from the historical tunneling data, the parameter trend detection model can be validated using the mud cake-related parameters in the test set, so that the hyperparameters of the parameter trend detection model can be adjusted based on the validation results.

[0065] The process described in the above embodiments is illustrated below through examples in practical applications.

[0066] The main methods for detecting sludge cake on the cutter head in related technologies are as follows:

[0067] Option 1: Install a cutterhead temperature detection device in the tunnel boring machine (TBM). This device includes a temperature acquisition module, a control module, a transmitting module, a receiving module, and a display module. The temperature acquisition module collects the real-time temperature of the cutterhead cutting surface and sends it to the control module. The control module determines the mud cake formation status based on the real-time temperature value and transmits the mud cake formation status to the receiving module via the transmitting module. The receiving module then sends the received mud cake formation status to the display module. The main component of the temperature acquisition module is a sensor that collects the temperature signal from the cutterhead cutting surface. This sensor operates in a harsh environment with strong vibrations, making it prone to damage. Furthermore, the addition of water to the soil chamber may affect the accuracy of the real-time temperature measurement. Additionally, this option only assesses the mud cake status based on the cutterhead temperature, resulting in a single assessment factor and an inability to accurately detect the mud cake formation status on the cutterhead.

[0068] Option 2: Collect key parameter data of the tunnel boring machine (TBM), divide the dataset into training and validation sets; set a target error; use a neural network to train the training set to obtain a cutterhead mud cake condition assessment model, input the validation set into the model to calculate the error rate; compare the error rate with the target error until the error rate is less than the target error; collect new key parameter data of the TBM, and use the cutterhead mud cake condition assessment model to perform real-time assessment and early warning of the cutterhead mud cake condition. This patent's data processing and analysis process is relatively complex. It uses a neural network to extract data features; as the number of neural network layers increases, the number of parameters increases, leading to computational complexity. Furthermore, it needs to consider the temperature of the excavated soil (which is easily affected by human factors), resulting in low accuracy in assessing the cutterhead mud cake condition.

[0069] Based on the defects in detecting cutterhead mud cake in the aforementioned related technologies, this application provides a cutterhead mud cake discrimination scheme based on CNN and LSTM. This scheme processes and analyzes tunnel boring machine data by combining one-dimensional convolutional layer Conv1D with LSTM, and judges the changing trends of cutterhead torque, total thrust, and thrust speed. It does not need to consider the temperature of the slag soil, and obtains a cutterhead mud cake discrimination model to achieve intelligent tunneling.

[0070] In this embodiment, the cutterhead torque, cutterhead speed, penetration depth, propulsion speed, and total propulsion force are used as input variables based on the tunneling data of the tunnel boring machine. The changing trends of each variable, such as cutterhead torque, propulsion speed, and total propulsion force, over a period of time are used as output variables. If the cutterhead torque continues to increase, the propulsion speed continues to decrease, and the total propulsion force continues to increase within the same time period, it indicates that the cutterhead is in a mud cake state during that period, and an early warning is issued. Otherwise, it is determined that the cutterhead is not in a mud cake state.

[0071] This embodiment can be divided into a model training stage and a mud cake detection stage.

[0072] The model training phase may include the following steps:

[0073] Step A1: Collect and filter the tunneling data of the tunnel boring machine.

[0074] Step A2: Normalize and segment the filtered dataset.

[0075] Step A3: Determine the network structure of the parameter trend detection model, which mainly includes two one-dimensional convolutional layers, two LSTM layers, and one fully connected layer.

[0076] Step A4: Determine the hyperparameters of the parameter trend detection model; the hyperparameters include: the number of neurons, batch size, activation function, loss function, etc.

[0077] Step A5: Input data for training, save the trained parameter trend detection model, and test the saved parameter trend detection model using the test set.

[0078] During the cake formation detection stage, a parameter trend detection model can be embedded into the control system. By analyzing and judging the changing trends of cutterhead torque, total propulsion force, and propulsion speed over a period of time, it can be determined whether the cutterhead is in a cake formation state.

[0079] Please see Figure 3 , Figure 3 The flowchart provided in this application, illustrating the training and testing of a parameter trend detection model using tunnel boring machine (TBM) data, specifically includes the following steps: Sensors collect data; PCC (Pearson Correlation Coefficient) is used to analyze the data in the PLC controller and industrial computer to filter key data, namely, mud cake-related parameters. The mud cake-related parameters are normalized and segmented to obtain training and testing sets. After building the model, the parameter trend detection model is trained and tested, and the model's effectiveness is evaluated.

[0080] The parsed tunnel boring machine (TBM) data file is large in volume, resulting in high computational costs for direct calculations, necessitating the removal of some data features. The Pearson Correlation Coefficient (PCC) method is used to extract input parameters highly correlated with cutterhead cake buildup. Data analysis reveals a strong correlation between cutterhead rotation speed, cutterhead torque, penetration depth, propulsion speed, and total propulsion force and cutterhead cake buildup. Therefore, this embodiment can select cutterhead rotation speed, cutterhead torque, penetration depth, propulsion speed, and total propulsion force for analysis within a specific time period (i.e., a historical timeframe).

[0081] The values ​​of cutterhead torque, cutterhead speed, penetration, feed rate, and total feed force after filtering have a large range. To facilitate data processing, all variables are normalized to the range [0,1]. The conversion formula is as follows:

[0082]

[0083] In the formula above, y represents the value after normalization, and x represents the current value. max This indicates the maximum range value of the column containing the current value.

[0084] The normalized data can be divided into training and testing sets in a 4:1 ratio.

[0085] The parameter trend detection model consists of two parts: one-dimensional convolutional processing and LSTM processing. The one-dimensional convolutional processing part mainly uses two one-dimensional convolutional layers (Conv-1D) to extract local features of the changes in propulsion speed, total propulsion force, and cutter head torque over a period of time. The LSTM processing part mainly uses two LSTM layers to capture the long-term trend features of the changes in propulsion speed, total propulsion force, and cutter head torque. The two LSTM layers compress the data into highly "condensed" data. Adding more layers can only lead to information loss during the information compression process and gradient vanishing during the training process.

[0086] This embodiment allows inputting propulsion speed, total propulsion force, cutterhead torque, cutterhead rotation speed, and penetration depth into the parametric trend detection model. Inputting cutterhead rotation speed and penetration depth into the model helps obtain more data features related to cutterhead mud cake formation during training. Combining one-dimensional convolutional layers and LSTM allows for more comprehensive and thorough extraction of features hidden within the tunnel boring machine (TBM) data. The one-dimensional convolutional layer and LSTM layer have 128 neurons, and the softsign activation function is chosen. The softsign activation function learns more efficiently and solves the gradient vanishing problem better than the tanh function. The concatenate function is used to combine the results of the two operations as input to the fully connected layer. The fully connected layer has 3 neurons and is mainly used to output the changing trends of the three variables: cutterhead torque, propulsion speed, and total propulsion force.

[0087] The objective function loss of the above parameter trend detection model is as follows:

[0088]

[0089] in and These are the results after processing by a one-dimensional convolutional layer and an LSTM layer, respectively. k and s k These are the true values, N represents the total number of samples, and λ||W||2 represents the regularization term, which is used to prevent the model from overfitting.

[0090] Specifically, a one-dimensional convolutional layer can output the duration of a trend over a short period of time (i.e., the first time period mentioned above). and slope The corresponding actual duration value is l k1 The true slope value is s k1 The LSTM layer can output the duration of a trend over a long period (i.e., the second time period mentioned above). and slope The corresponding actual duration value is l k2 The true slope value is s k2 .

[0091] Therefore, the objective function is:

[0092] LOSS = loss1 + loss2;

[0093]

[0094]

[0095] λ1||W1||2 represents the regularization term corresponding to the training process of the one-dimensional convolutional layer, and λ2||W2||2 represents the regularization term corresponding to the training process of the LSTM.

[0096] Set the hyperparameters for training, and then validate the model using a test set. During training, observe the loss and accuracy curves, and fine-tune the hyperparameters. When the loss curve stabilizes and the accuracy value reaches its maximum and stabilizes, save the relevant parameters and model file.

[0097] Please see Figure 4 , Figure 4 The flowchart provided in this application embodiment illustrates a model transfer to a tunnel boring machine (TBM) control system. The specific process includes: calling the model file of a parameter trend detection model; reading the model file; inputting the total thrust, cutterhead torque, and thrust speed values ​​over a period of time; and outputting the changing trends of the total thrust, cutterhead torque, and thrust speed over that period. If the trends match the characteristics of the three parameters when the cutterhead forms mud cake, an early warning is issued; otherwise, normal tunneling continues.

[0098] This embodiment uses a combination of one-dimensional convolutional layers and LSTM to process tunnel boring machine (TBM) data, judging the trends of cutterhead torque, propulsion speed, and total propulsion force during tunneling. This embodiment implements an intelligent cutterhead mud cake detection system based on one-dimensional convolutional layers and LSTM, achieving real-time determination of whether the cutterhead is in a mud cake state. The above embodiment can intelligently determine whether the cutterhead is in a mud cake state, avoiding severe cutterhead wear, extending service life, and improving excavation efficiency. The above embodiment reduces computational complexity, shortens computation time, and lowers the requirements for computer hardware.

[0099] Please see Figure 5 , Figure 5 A schematic diagram of the structure of a device for detecting sludge cake on a cutter head, provided in an embodiment of this application;

[0100] The device may include:

[0101] The parameter extraction module 501 is used to read historical tunneling data from the PLC controller and industrial control computer of the tunnel boring machine, and extract mud cake-related parameters from the historical tunneling data;

[0102] The model training module 502 is used to train a parameter trend detection model using the mud cake-related parameters in the historical tunneling data; wherein, the parameter trend detection model includes a one-dimensional convolutional layer, an LSTM layer, and a fully connected layer;

[0103] The trend detection module 503 is used to obtain the mud cake-related parameters of the tunnel boring machine in the target time period, input the mud cake-related parameters in the target time period into the parameter trend detection model, and obtain the changing trend of the mud cake-related parameters in the target time period.

[0104] The judgment module 504 is used to determine whether the changing trend of the mud cake-related parameters within the target time period conforms to the characteristics of cutterhead mud cake; if yes, it determines that the tunnel boring machine is in the state of cutterhead mud cake; if no, it determines that the tunnel boring machine is not in the state of cutterhead mud cake.

[0105] This embodiment extracts mud cake-related parameters from historical tunneling data of the tunnel boring machine (TBM) and trains a parameter trend detection model using these parameters. This model outputs the changing trends of these parameters. After inputting the mud cake-related parameters for a target time period into the model, it outputs the changing trends within that period. Since these parameters are related to the mud cake state of the TBM cutterhead, the model's output trend can be used to evaluate whether the TBM is experiencing cutterhead mud cake formation. This process, detecting cutterhead mud cake formation based on the changing trends of these parameters, reduces interference from the working environment and improves the accuracy of cutterhead mud cake detection.

[0106] Furthermore, the process by which the parameter extraction module 501 extracts mud cake-related parameters from the historical tunneling data includes:

[0107] Extract mud cake-related parameters from the historical tunneling data that have a correlation greater than a preset value with the state of mud cake on the cutterhead; wherein, the mud cake-related parameters include propulsion speed, total propulsion force, and cutterhead torque.

[0108] Furthermore, the parameters related to the sludge cake formation also include the cutter head rotation speed and penetration depth.

[0109] Furthermore, the process by which the model training module 502 trains the parameter trend detection model using the mud cake-related parameters in the historical tunneling data includes: inputting the mud cake-related parameters in the historical tunneling data into the one-dimensional convolutional layer to obtain the local trend features of the mud cake-related parameters; inputting the mud cake-related parameters in the historical tunneling data into the LSTM layer to obtain the long-term trend features of the mud cake-related parameters; and training the parameter trend detection model using the local trend features, the long-term trend features, and the objective function; wherein the local trend features are used to describe the change trend within a first time period, the long-term trend features are used to describe the change trend within a second time period, the first time period is shorter than the second time period, and the change trend includes the trend duration and slope.

[0110] Furthermore, the process by which the model training module 502 trains the parameter trend detection model using the mud cake-related parameters in the historical tunneling data includes: extracting a training set from the historical tunneling data according to a preset ratio; normalizing the mud cake-related parameters in the training set according to the maximum range value of the mud cake-related parameters; and training the parameter trend detection model using the normalized training set.

[0111] Furthermore, it also includes:

[0112] The alarm module is used to control the tunnel boring machine to stop tunneling and output alarm information after determining that the tunnel boring machine is in a state of mud cake formation on the cutterhead.

[0113] Furthermore, if the parameters related to cake formation include propulsion speed, total propulsion force, and cutterhead torque, then the process by which the judgment module 504 judges whether the changing trends of the parameters related to cake formation within the target time period conform to the characteristics of cutterhead cake formation includes: judging whether the changing trends of propulsion speed, total propulsion force, and cutterhead torque within the target time period conform to the characteristics of cutterhead cake formation; wherein, the characteristics of cutterhead cake formation are that the propulsion speed continuously decreases, the total propulsion force continuously increases, and the cutterhead torque continuously increases within the target time period.

[0114] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0115] This application also provides a storage medium on which a computer program is stored, which, when executed, can perform the steps provided in the above embodiments. The storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0116] This application also provides an electronic device that may include a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it can implement the steps provided in the above embodiments. Of course, the electronic device may also include various network interfaces, power supplies, and other components.

[0117] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0118] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for detecting mud cake formation on a cutter head, characterized in that, include: Historical tunneling data is read from the PLC controller and industrial computer of the tunnel boring machine, and parameters related to mud cake are extracted from the historical tunneling data; A parameter trend detection model is trained using the mud cake-related parameters in the historical tunneling data; wherein, the parameter trend detection model includes a one-dimensional convolutional layer, an LSTM layer, and a fully connected layer; The relevant parameters of mud cake formation of the tunnel boring machine within the target time period are obtained, and the relevant parameters of mud cake formation within the target time period are input into the parameter trend detection model to obtain the changing trend of the relevant parameters of mud cake formation within the target time period. Determine whether the changing trend of the mud cake-related parameters within the target time period conforms to the characteristics of cutterhead mud cake; if yes, determine that the tunnel boring machine is in the state of cutterhead mud cake; if no, determine that the tunnel boring machine is not in the state of cutterhead mud cake. The training of a parameter trend detection model using mud cake-related parameters from the historical tunneling data includes: A training set is extracted from the historical tunneling data according to a preset ratio; The parameters related to mud cake in the training set are normalized based on the maximum range value of the parameters related to mud cake. The mud cake-related parameters from the normalized training set are input into the one-dimensional convolutional layer to obtain the local trend features of the mud cake-related parameters. The parameters related to mud cake in the normalized training set are input into the LSTM layer to obtain the long-term trend characteristics of the parameters related to mud cake. The parametric trend detection model is trained using the local trend features, the long-term trend features, and the objective function. The local trend feature is used to describe the trend of change within a first time period, and the long-term trend feature is used to describe the trend of change within a second time period. The first time period is shorter than the second time period, and the trend of change includes the trend duration and slope.

2. The method for detecting clay cake on a cutter head according to claim 1, characterized in that, Extract parameters related to mud cake from the historical tunneling data, including: Extract mud cake-related parameters from the historical tunneling data that have a correlation greater than a preset value with the state of mud cake on the cutterhead; wherein, the mud cake-related parameters include propulsion speed, total propulsion force, and cutterhead torque.

3. The method for detecting clay cake on a cutter head according to claim 2, characterized in that, The parameters related to the mud cake formation also include the cutter head rotation speed and penetration depth.

4. The method for detecting clay cake on a cutter head according to claim 1, characterized in that, After determining that the tunnel boring machine is in a state of mud cake formation on the cutterhead, the process further includes: The tunnel boring machine is controlled to stop tunneling and an alarm message is output.

5. The method for detecting clay cake on a cutter head according to any one of claims 1 to 4, characterized in that, If the parameters related to cake formation include propulsion speed, total propulsion force, and cutterhead torque, then it is determined whether the changing trend of the parameters related to cake formation within the target time period conforms to the characteristics of cutterhead cake formation, including: Determine whether the changing trends of propulsion speed, total propulsion force, and cutterhead torque within the target time period conform to the characteristics of cutterhead cake formation. The characteristics of the cutterhead forming mud cake are that the propulsion speed continuously decreases, the total propulsion force continuously increases, and the cutterhead torque continuously increases during the target time period.

6. A device for detecting mud cake formation on a cutter head, characterized in that, include: The parameter extraction module is used to read historical tunneling data from the PLC controller and industrial control computer of the tunnel boring machine, and extract mud cake-related parameters from the historical tunneling data; The model training module is used to train a parameter trend detection model using the mud cake-related parameters in the historical tunneling data; wherein, the parameter trend detection model includes a one-dimensional convolutional layer, an LSTM layer, and a fully connected layer; The trend detection module is used to obtain the mud cake-related parameters of the tunnel boring machine within a target time period, input the mud cake-related parameters within the target time period into the parameter trend detection model, and obtain the changing trend of the mud cake-related parameters within the target time period. The judgment module is used to determine whether the changing trend of the mud cake-related parameters within the target time period conforms to the characteristics of cutterhead mud cake; if yes, the tunnel boring machine is determined to be in the state of cutterhead mud cake; if no, the tunnel boring machine is determined not to be in the state of cutterhead mud cake. The process by which the model training module trains the parameter trend detection model using the mud cake-related parameters in the historical tunneling data includes: A training set is extracted from the historical tunneling data according to a preset ratio; The parameters related to mud cake in the training set are normalized based on the maximum range value of the parameters related to mud cake. The mud cake-related parameters from the normalized training set are input into the one-dimensional convolutional layer to obtain the local trend features of the mud cake-related parameters. The parameters related to mud cake in the normalized training set are input into the LSTM layer to obtain the long-term trend characteristics of the parameters related to mud cake. The parametric trend detection model is trained using the local trend features, the long-term trend features, and the objective function. The local trend feature is used to describe the trend of change within a first time period, and the long-term trend feature is used to describe the trend of change within a second time period. The first time period is shorter than the second time period, and the trend of change includes the trend duration and slope.

7. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program, and the processor, when calling the computer program in the memory, implements the steps of the method for detecting mud cake on the cutter head as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the steps of the method for detecting sludge cake on the cutter head as described in any one of claims 1 to 5.

Images