TBM cutterhead torque prediction method and device

Abstract

The invention discloses a TBM cutterhead torque prediction method and device. The method comprises the following steps that a first tunneling parameter related to TBM cutterhead torque is obtained; the linear and nonlinear correlation degree between the first tunneling parameter and the cutterhead torque is measured, and a second tunneling parameter is screened out; preprocessing the second tunneling parameter, and sequentially constructing a data set comprising a training set, a verification set and a test set according to a time sequence; constructing a fusion network model based on a space-time attention mechanism, a time convolution network and a bidirectional long-short term memory network; training a fusion network model by using the training set, adjusting model hyper-parameters based on the verification set, and checking the generalization performance of the model on the test set to obtain a trained fusion model; and predicting the TBM cutterhead torque by using the fusion model. According to the method, tunneling parameters are screened out in a mode of combining experience knowledge and correlation analysis, a novel network integrating multi-scale time convolution and a two-way long-short-term memory network under space-time attention is designed for torque accurate prediction, early warning can be provided for TBM jamming disasters, guidance is provided for tunneling parameter optimization, and the method has the advantages of being high in reliability and high in reliability. And the method has important engineering application value.

Description

TECHNICAL FIELD

[0001] The present application relates to a TBM cutterhead torque prediction method and device, belonging to the technical field of hard rock TBM engineering combined with machine learning. BACKGROUND

[0002] Hard rock tunnel boring machine (TBM) method has the advantages of safety, efficiency, environmental protection and other advantages compared with traditional drilling and blasting method, and is gaining more and more applications in traffic tunnel, water tunnel, mine roadway and other excavations. However, the adaptability of TBM to geology is poor, that is, the tunneling performance of TBM is very sensitive to geological changes, and the tunneling control parameters (total thrust and cutterhead speed) need to be adjusted in time according to the encountered geology to ensure the safe and efficient tunneling of TBM. TBM cutterhead torque is a load, passive parameter, and its trend is helpful for the optimization of tunneling control parameters. At the same time, the trend of TBM cutterhead torque can also reflect the stability of the rock mass in front of the working face, and has an indicative effect on the potential cutterhead jam risk. Therefore, accurate prediction of TBM cutterhead torque is of great significance for optimizing the tunneling control parameters and warning the cutterhead jam risk.

[0003] Recurrent neural networks, such as classic long short-term memory LSTM, gated recurrent unit GRU, etc., are common methods for time series prediction, and have been successfully applied in many fields, especially in short sequence prediction. However, they cannot effectively capture dependencies from long sequence data and do not support parallel computing. Time convolutional network TCN is a time series prediction method proposed in recent years, which overcomes the shortcomings of recurrent neural networks through causal convolution, dilated convolution and residual connection, and can effectively capture long sequence dependencies and support parallel computing. However, when TCN uses causal dilated convolution to extract features, it cannot distinguish the importance of features, which is not conducive to improving the prediction accuracy of the model.

[0004] Based on the above analysis, there is an urgent need for a TBM cutterhead torque prediction method with high prediction accuracy. SUMMARY

[0005] In view of the problem of low prediction accuracy of current TBM cutterhead torque, the present application provides a TBM cutterhead torque prediction method and device. The method of the present application constructs a fusion network model based on multi-scale time convolution and bidirectional long short-term memory network under spatiotemporal attention, which can accurately capture multi-scale long and short sequence dependencies, and efficiently identify key features by introducing spatiotemporal attention mechanism, finally realize the accurate and real-time prediction of TBM cutterhead torque.

[0006] In the first aspect, a TBM cutterhead torque prediction method is provided, comprising: obtaining a first tunneling parameter related to TBM cutterhead torque; measure the linear and nonlinear correlation degree between the first tunneling parameter and the cutterhead torque, and screen out the second tunneling parameter; The second tunneling parameter is preprocessed, and a data set including a training set, a validation set and a test set is constructed in time sequence; A fusion network model based on a space-time attention mechanism, a time convolution network and a bidirectional long short-term memory network is constructed. The fusion network model is trained using the training set, the model hyperparameters are adjusted based on the validation set, and the generalization performance of the model is verified on the test set to obtain the trained fusion model. The fusion model is used to predict the TBM cutterhead torque.

[0007] In a possible implementation manner, the first tunneling parameter includes: cutterhead torque, cutterhead speed, total thrust, net tunneling speed, TBM overturning angle, support shoe tensioning force, main drive motor 1 torque, main drive motor 1 current, main drive motor 1 speed, main drive motor 1 temperature, 1 # top shield cylinder rodless cavity pressure, 2 # top shield cylinder rodless cavity pressure, left shield cylinder rodless cavity pressure, left wedge block cylinder rodless cavity pressure, right shield cylinder rodless cavity pressure, right wedge block cylinder rodless cavity pressure, propulsion cylinder rodless cavity pressure, support shoe support pressure, left torque oil cylinder rodless cavity pressure, right torque oil cylinder rodless cavity pressure, left support shoe front and rear pitch angle, left support shoe up and down overturning angle, right support shoe front and rear pitch angle, right support shoe up and down overturning angle, propulsion pump pressure, support shoe pump pressure, auxiliary pump pressure, main belt conveyor drive pressure, propulsion pump motor current, support shoe pump motor current, main belt conveyor pump motor current and rear auxiliary belt conveyor current.

[0008] In a possible implementation manner, the measuring the linear and nonlinear correlation degree between the first tunneling parameter and the cutterhead torque includes: The linear and nonlinear correlation degree between the first tunneling parameter and the cutterhead torque is measured by using the Pearson correlation coefficient PLCC and the maximum information coefficient MIC, and the tunneling parameter with the absolute value of PLCC greater than or equal to 0.3 and MIC greater than or equal to 0.3 is screened out. Considering the symmetry of mechanical design, the tunneling parameter of the symmetric tunneling parameter with the absolute value of PLCC less than 0.3 or MIC less than 0.3 is further screened out from the data excluded from the first tunneling parameter as the second tunneling parameter.

[0009] In a possible implementation manner, the preprocessing includes data cleaning and data normalization.

[0010] In a possible implementation manner, the fusion model network includes: A stacked convolution layer is used for feature extraction of short, medium and long three scales of time sequences in the original time sequence in the data set. a time convolution network, configured to extract features of three scales of the time sequence and splice the features to obtain first time sequence features; a spatio-temporal attention mechanism, configured to determine importance of each feature in the first time sequence features and assign a weight to the feature to obtain an attention-weighted feature; a Bi-LSTM layer, configured to capture context relationships of the input attention-weighted features and output to a full connection layer to realize the prediction of the cutterhead torque.

[0011] Further, the stacked convolution layer includes a convolution layer, a normalization layer and an activation function; different convolution kernel sizes are set for different convolution layers to process time sequences of different scales.

[0012] In a second aspect, a TBM cutterhead torque prediction device is provided, including: an acquisition module, configured to acquire first tunneling parameters related to the TBM cutterhead torque; a screening module, configured to measure linear and nonlinear correlation degrees between the first tunneling parameters and the cutterhead torque and screen second tunneling parameters; a preprocessing module, configured to preprocess the second tunneling parameters and sequentially construct a data set including a training set, a validation set and a test set according to a time sequence; a construction module, configured to construct a fusion network model based on a spatio-temporal attention mechanism, a time convolution network and a bidirectional long short-term memory network; a training module, configured to train the fusion network model by using the training set, adjust model hyperparameters based on the validation set, verify a generalization performance of the model on the test set and obtain a trained fusion model; an output module, configured to predict the TBM cutterhead torque by using the fusion model.

[0013] In a third aspect, an electronic device is provided, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the TBM cutterhead torque prediction method of the first aspect when executing the program.

[0014] In a fourth aspect, a non-transitory computer readable storage medium is provided, having a computer program stored thereon, and the computer program is executable by a processor to implement the TBM cutterhead torque prediction method of the first aspect.

[0015] In a fifth aspect, a computer program product is provided, including a computer program, and the computer program is executable by a processor to implement the TBM cutterhead torque prediction method of the first aspect.

[0016] Compared with existing technologies, this method offers the following advantages: Compared to recurrent neural networks, it effectively captures dependencies between long sequences of data and enables parallel computation of most network layers. Compared to conventional temporal convolutional networks, it can discriminate the importance of extracted features, resulting in higher prediction accuracy. This method can provide early warnings for TBM jams and guidance for optimizing tunneling parameters, possessing significant engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0018] Figure 1 Flowchart of the TBM torque prediction method integrating spatiotemporal attention multi-scale temporal convolution with bidirectional long short-term memory network provided by an embodiment of the present invention; Figure 2 The cutterhead torque prediction results of the embodiment of the present invention on the validation set and two test sets are shown; Figure 3 A schematic structural diagram of a TBM cutterhead torque prediction device provided by an embodiment of the present invention; Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the accompanying drawings. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort shall fall within the scope of protection of the present invention.

[0020] refer to Figure 1 , is a TBM cutterhead torque prediction method, comprising the following steps: S101. Obtain a first tunneling parameter related to the TBM cutterhead torque.

[0021] In a possible implementation manner, the first tunneling parameters include: cutter head torque, cutter head rotating speed, total thrust, net tunneling speed, TBM overturning angle, support shoe bracing force, main drive motor 1 torque, main drive motor 1 current, main drive motor 1 rotating speed, main drive motor 1 temperature, 1 # top shield cylinder rodless cavity pressure, 2 # top shield cylinder rodless cavity pressure, left shield cylinder rodless cavity pressure, left wedge cylinder rodless cavity pressure, right shield cylinder rodless cavity pressure, right wedge cylinder rodless cavity pressure, advancing cylinder rodless cavity pressure, support shoe support pressure, left torque cylinder rodless cavity pressure, right torque cylinder rodless cavity pressure, left support shoe front-rear pitch angle, left support shoe up-down overturning angle, right support shoe front-rear pitch angle, right support shoe up-down overturning angle, advancing pump pressure, support shoe pump pressure, auxiliary pump pressure, main belt conveyor drive pressure, advancing pump motor current, support shoe pump motor current, main belt conveyor pump motor current, and rear auxiliary belt conveyor current.

[0022] In a possible implementation manner, the first tunneling parameters are filtered from original data by a CSM model (Colorado School of Mines) and experience analysis.

[0023] S102, measure the linear and nonlinear correlation degree between the first tunneling parameter and the cutter head torque, and filter out a second tunneling parameter.

[0024] In a possible implementation manner, S102 includes: S102a, measure the linear and nonlinear correlation degree between the first tunneling parameter and the cutter head torque by using a Pearson correlation coefficient PLCC and a maximum information coefficient MIC, and filter out a tunneling parameter whose absolute value of PLCC is greater than or equal to 0.3 and whose MIC is greater than or equal to 0.3; S102b, considering the symmetry of mechanical design, further filter out, from the data filtered out from the first tunneling parameter, a symmetric tunneling parameter whose absolute value of PLCC is less than 0.3 or whose MIC is less than 0.3, as the second tunneling parameter.

[0025] It should be noted that the symmetry of mechanical design refers to components of the TBM in a symmetric structure, such as the left torque cylinder rodless cavity and the right torque cylinder rodless cavity. For example, S102a has filtered out, by correlation analysis, left side cylinder tunneling parameters whose absolute value of PLCC is greater than or equal to 0.3 and whose MIC is greater than or equal to 0.3, however, it is found in the analysis process that sometimes the left side cylinder parameters meet the range while the right side symmetric cylinder parameters do not meet the range, and in fact the right side cylinder parameters should also be considered to be included in the second tunneling parameters.

[0026] Further, the second tunneling parameters include: cutterhead torque, cutterhead rotating speed, total thrust, net tunneling speed, TBM overturning angle, main drive motor 1 torque, main drive motor 1 current, main drive motor 1 rotating speed, 1# top shield oil cylinder rodless cavity pressure, 2# top shield oil cylinder rodless cavity pressure, left wedge block cylinder rodless cavity pressure, right wedge block cylinder rodless cavity pressure, advancing cylinder rodless cavity pressure, left support shoe up-down overturning angle, right support shoe up-down overturning angle, advancing pump pressure, support shoe pump pressure, advancing pump motor current, main belt conveyor pump motor current and rear auxiliary belt conveyor current.

[0027] S103, preprocessing the second tunneling parameters to sequentially construct a data set including a training set, a validation set and a test set in time sequence.

[0028] In a possible implementation manner, the preprocessing in S103 includes data cleaning and data normalization.

[0029] It should be noted that the time sequence refers to processing data in the order of collection time, and the data is not scattered.

[0030] Further, the data cleaning includes: (1) filtering out data in the tunneling state (determined by the machine state code) from the second tunneling data; (2) deleting data points with total thrust, cutterhead rotating speed, net tunneling speed or torque of 0; (3) deleting data points with net tunneling speed exceeding 120 mm / min (exceeding the sensor range); (4) further detecting and deleting outliers by using the Isolation Forest algorithm.

[0031] Further, the data normalization includes normalization based on the sensor range of the second tunneling data.

[0032] Further, the data set has features of tunneling parameters filtered and processed at a moment in a historical N time step data segment (for example, 20xN in shape), and a label of cutterhead torque at the next moment of the moment (1 in shape).

[0033] It should be noted that the torque is a single output result, so the shape is 1. For example, if the output is torque and thrust, the output data shape is 2.

[0034] S104, constructing a fusion network model (CBAM-MTCN-BiLSTM) based on a space-time attention mechanism, a time convolution network and a bidirectional long short-term memory network.

[0035] In a possible implementation manner, referring to Figure 2 , the fusion network model includes: stacked convolutional layers for feature extraction of short, medium and long scale time series in the original time series in the dataset; temporal convolutional network for feature extraction of three scales of time series and concatenation to obtain first time series features; spatio-temporal attention mechanism for determining the importance of each feature in the first time series features and assigning weights to obtain attention-weighted features; Bi-LSTM layer for capturing the context relationship of the input attention-weighted features, and outputting to a fully connected layer to realize the prediction of cutter torque.

[0036] Further, the stacked convolutional layers include convolutional layers, normalization layers and activation functions; different convolutional kernel sizes are set to process time series of different scales.

[0037] For example, the stacked convolutional layers include Conv+BN+ReLU, and the same original sequence is input to three stacked convolutional layers (which can be regarded as filters) to obtain short, medium and long features. Among them, the short time series Conv corresponds to a convolutional kernel size of 1, the medium time series is not processed, and the long time series stacked Conv has a convolutional kernel size of 3.

[0038] Further, the temporal convolutional network (TCN) includes 3 TCN layers, each corresponding to a scale. The three TCN layers have the same number of filters (convenient for subsequent concatenation) and different dilation rates (used to extract features of different receptive fields).

[0039] Further, the spatio-temporal attention mechanism (CBAM) is a lightweight attention mechanism for convolutional neural networks, which dynamically enhances important information and suppresses irrelevant information in the feature map through the series structure of channel attention and spatial attention modules.

[0040] Further, the Bi-LSTM layer is used to capture the context relationship of the input attention-weighted features, and the output is a fully connected layer to realize the prediction of cutter torque.

[0041] The inference process of the fusion network model is as follows: first, one-dimensional stacked convolutional layers (Conv+BN+ReLU) are used to extract short, medium and long scale features from the original time series, each scale corresponding to a TCN layer, and the three TCN layers having the same number of filters (convenient for subsequent concatenation) and different dilation rates (used to extract features of different receptive fields); second, the features extracted under the three scales (feature 1, feature 2 and feature 3) are concatenated, and the spatio-temporal attention mechanism layer is used to determine the importance of each feature, so that the model pays more attention to important features; finally, the attention-weighted features are input into the multi-layer Bi-LSTM, and the output layer is a fully connected layer (dimension 1) to realize the prediction of cutter torque.

[0042] S105, training the fusion network model by using the training set, adjusting the model hyperparameters based on the validation set, and verifying the generalization performance of the model on the test set to obtain the trained fusion model.

[0043] In a possible implementation manner, the adjusting the model hyperparameters based on the validation set includes the number of convolution layers and the size of convolution kernels, the number of TCN layers, the size of convolution kernels, the dropout rate, the number of units of the Bi-LSTM layer, and the like. When adjusting the hyperparameters of a certain module, the hyperparameters of other modules remain unchanged. Whether the hyperparameters are appropriate is determined by the prediction performance of the model on the validation set. Finally, the optimal combination of hyperparameters is obtained.

[0044] S106, predicting the TBM cutterhead torque by using the fusion model.

[0045] Figure 2 The models (LSTM, GRU, Bi-LSTM, Bi-GRU, TCN, and CBAM-MTCN-BiLSTM (CMTB, the model of the embodiment)) provided by the embodiment of the present application perform on the validation set and the test set. It can be seen from Figure 2 that the prediction accuracy of the model is the highest on the two test sets, and the R 2 values are 0.709 and 0.712, respectively.

[0046] The TBM cutterhead torque prediction device provided by the present application is described below. The TBM cutterhead torque prediction device described below can be correspondingly referred to the TBM cutterhead torque prediction method described above.

[0047] Figure 3 is a structural schematic diagram of the TBM cutterhead torque prediction device provided by the embodiment of the present application, as shown in Figure 3 includes an acquisition module 31, a screening module 32, a preprocessing module 33, a construction module 34, a training module 35, and an output module 36, wherein: The acquisition module 31 is configured to acquire first tunneling parameters related to the TBM cutterhead torque. The screening module 32 is configured to measure the linear and nonlinear correlation degree between the first tunneling parameters and the cutterhead torque, and screen out second tunneling parameters. The preprocessing module 33 is configured to preprocess the second tunneling parameters, and sequentially construct a data set including a training set, a validation set, and a test set according to a time sequence. The construction module 34 is configured to construct a fusion network model based on a spatiotemporal attention mechanism, a time convolution network, and a bidirectional long short-term memory network. The training module 35 is configured to train the fusion network model by using the training set, adjust the model hyperparameters based on the validation set, and verify the generalization performance of the model on the test set to obtain the trained fusion model. The output module 36 is configured to predict the TBM cutterhead torque by using the fusion model.

[0048] Figure 4 An example of an entity structure diagram of an electronic device is shown in Figure 4 As shown, the electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 can communicate with each other through the communications bus 440. The processor 410 can invoke the logical instructions in the memory 430 to execute the TBM cutterhead torque prediction method.

[0049] In addition, the logical instructions in the memory 430 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, includes several instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0050] On the other hand, the present application also provides a computer program product, which includes a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program is executed by a processor, so that the computer can execute the TBM cutterhead torque prediction method provided by the above-mentioned methods.

[0051] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, which is executed by a processor to implement the TBM cutterhead torque prediction method provided by the above-mentioned methods.

[0052] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0053] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0054] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. TBM cutterhead torque prediction method, characterized in that: include: Obtaining a first excavation parameter related to the TBM cutterhead torque; Measuring the linear and nonlinear correlation between the first excavation parameter and the cutterhead torque, and selecting a second excavation parameter; Preprocessing the second excavation parameters, and constructing a data set including a training set, a validation set, and a test set in chronological order; Construct a fusion network model based on spatiotemporal attention mechanism, temporal convolutional network and bidirectional long short-term memory network; Use the training set to train the fusion network model and adjust the model hyperparameters based on the validation set. Then, test the generalization performance of the model on the test set to obtain the trained fusion model. Predicting TBM cutterhead torque using fusion model.

2. The TBM cutterhead torque prediction method according to claim 1, wherein: The first tunneling parameter includes: cutter head torque, cutter head speed, total thrust, net tunneling speed, TBM flip angle, gripper clamping force, main drive motor 1 torque, main drive motor 1 current, main drive motor 1 speed, main drive motor 1 temperature, 1# top shield cylinder rodless chamber pressure, 2# top shield cylinder rodless chamber pressure, left shield cylinder rodless chamber pressure, left wedge cylinder rodless chamber pressure, right shield cylinder rodless chamber pressure, right wedge cylinder rodless chamber pressure, thrust cylinder rodless chamber pressure, gripper support pressure, left torque cylinder rodless chamber pressure, right torque cylinder rodless chamber pressure, left gripper front and rear pitch angle, left gripper up and down flip angle, right gripper front and rear pitch angle, right gripper up and down flip angle, thrust pump pressure, gripper pump pressure, auxiliary pump pressure, main belt conveyor drive pressure, thrust pump motor current, gripper pump motor current, main belt conveyor pump motor current and rear supporting belt conveyor current.

3. The TBM cutterhead torque prediction method according to claim 1, characterized in that: The measuring of the linear and nonlinear correlation between the first tunneling parameter and the cutterhead torque includes: The Pearson correlation coefficient (PLCC) and maximum information coefficient (MIC) were used to measure the linear and nonlinear correlation between the first tunneling parameter and the cutterhead torque, and tunneling parameters with an absolute value of PLCC greater than or equal to 0.3 and a MIC greater than or equal to 0.3 were selected. Considering the symmetry of the mechanical design, excavation parameters with symmetrical excavation parameters whose absolute value of PLCC is less than 0.3 or MIC is less than 0.3 are further selected from the data eliminated from the first excavation parameters as the second excavation parameters.

4. The TBM cutterhead torque prediction method according to claim 1, characterized in that: The preprocessing includes: data cleaning and data normalization.

5. The TBM cutterhead torque prediction method according to claim 1, characterized in that: The fusion model network includes: Stacked convolutional layers are used to extract features of short, medium, and long time series from the original time series in the dataset; Temporal convolutional network, used to extract the features of time series at three scales and concatenate them to obtain the first time series features; The spatiotemporal attention mechanism is used to determine the importance of each feature in the first time series feature and assign weights to obtain attention-weighted features; The Bi-LSTM layer is used to capture the contextual relationship of the input intention-weighted features, and the output is a fully connected layer to predict the cutterhead torque.

6. The TBM cutterhead torque prediction method according to claim 5, characterized in that: The stacked convolutional layer includes a convolutional layer, a normalization layer and an activation function; by setting the convolution kernel size of different convolutional layers, time series of different scales can be processed.

7. TBM cutter head torque prediction device, characterized in that: include: An acquisition module, configured to acquire a first excavation parameter related to the TBM cutterhead torque; a screening module, configured to measure the linear and nonlinear correlation between the first tunneling parameter and the cutterhead torque, and screen out a second tunneling parameter; A preprocessing module is used to preprocess the second excavation parameter and construct a data set including a training set, a validation set and a test set in chronological order; A building block for constructing a fusion network model based on spatiotemporal attention mechanism, temporal convolutional network, and bidirectional long short-term memory network; The training module is used to train the fusion network model using the training set, adjust the model hyperparameters based on the validation set, and test the generalization performance of the model on the test set to obtain the trained fusion model; Output module, used to predict TBM cutterhead torque using the fusion model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the TBM cutter head torque prediction method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the TBM cutterhead torque prediction method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the TBM cutterhead torque prediction method according to any one of claims 1 to 6 is implemented.

Images