Tunneling parameter prediction method, system and equipment of tunnel boring machine
By combining feature importance ranking with neural network models, the problem of accuracy in predicting tunnel boring machine parameters was solved, enabling the capture of sudden local changes and overall trends, thereby improving construction efficiency and safety.
Patent Information
- Application Number
- CN202510872341.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-26
- Publication Date
- 2025-11-21
AI Technical Summary
Existing methods for predicting the parameters of tunnel boring machines rely on a single model, which makes it difficult to effectively capture the sudden local changes and overall trends of the equipment in complex geological environments. This results in poor prediction accuracy and affects construction efficiency and safety.
The feature set is refined by ranking features based on their importance. Combined with long short-term memory network and convolutional neural network model, local variation features are extracted by wavelet transform and then weighted and fused to achieve accurate prediction of tunnel excavation parameters.
It improves the accuracy of tunnel boring machine parameter prediction and the robustness of the model, optimizes the tunneling process, and enhances construction efficiency and safety.
Smart Images
Figure CN120995252A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of civil engineering technology, specifically to a method, system, and equipment for predicting tunneling parameters of a tunnel boring machine. Background Technology
[0002] In related technologies, the selection of input feature parameters for tunnel boring machine (TBM) parameter prediction often relies on experience, making it susceptible to subjective judgment and lacking a deep understanding of the equipment's dynamic behavior in complex geological environments. This hinders efficient, accurate, and adaptive control. TBM construction parameter prediction methods primarily depend on single models, such as LSTM (Long Short-Term Memory) or CNN (Convolutional Neural Network) models. However, single models have limitations in handling complex local rock-machine nonlinear relationships and capturing long-term trends, leading to poor accuracy in predictions under certain geological conditions, such as relatively complex geological conditions. Summary of the Invention
[0003] In view of the above-mentioned defects or deficiencies in the existing technology, it is desirable to provide a method, system and equipment for predicting tunneling parameters of tunnel boring machines, which can effectively capture the local sudden changes and overall trend evolution of tunnel boring machine operating parameters, improve prediction accuracy and model robustness, and thus optimize the tunneling process and improve construction efficiency and safety.
[0004] In a first aspect, embodiments of this application provide a method for predicting tunneling parameters of a tunnel boring machine, including:
[0005] The importance of features is ranked according to the tunneling parameters of the tunnel boring machine, and a refined feature set is selected based on the ranking results.
[0006] The refined feature set is input into a pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step.
[0007] Wavelet transform is applied to the refined feature set to extract local variation features from the refined feature set;
[0008] By inputting global trend and local change features into a pre-trained convolutional neural network model, a second prediction result of the tunneling parameters at the next time step is obtained;
[0009] By combining the first prediction result and the second prediction result, the final prediction result of the tunneling parameters at the next moment is obtained.
[0010] In some examples, the process of ranking features by importance based on the tunneling parameters of the tunnel boring machine and selecting a refined feature set based on the ranking results includes:
[0011] Obtain the tunneling parameters of the tunnel boring machine;
[0012] The tunneling parameters of the tunnel boring machine are standardized to obtain the standardized characteristics of each tunneling parameter;
[0013] Evaluate the information gain of the standardized features of each tunneling parameter across all tree species;
[0014] The importance of the standardized features of each tunneling parameter is ranked according to the information gain.
[0015] The refined feature set is obtained based on the standardized features of the N most important tunneling parameters, where N is less than the number of types of tunneling parameters.
[0016] In some examples, the process of filtering the refined feature set based on the sorting results specifically includes:
[0017] The tunneling parameters corresponding to the refined feature set are constructed as time-series data in the form of a sliding window in chronological order;
[0018] The refined feature set is obtained based on the time series data.
[0019] In some examples, performing wavelet transform on the refined feature set to extract local variation features from the refined feature set includes:
[0020] Multi-scale wavelet decomposition is used to perform wavelet transform operations on the features in the refined feature set to obtain approximation coefficients and detail coefficients at multiple scales.
[0021] Based on the approximation coefficients and detail coefficients at the multiple scales, a wavelet feature matrix is formed, wherein the wavelet feature matrix characterizes the local variation features.
[0022] In some examples, fusing the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment includes:
[0023] The first prediction result and the second prediction result are concatenated, and the weighting coefficients are obtained based on the concatenation result;
[0024] Based on the weighting coefficients, the first prediction result and the second prediction result are weighted and fused to obtain the fused feature;
[0025] The final prediction result of the tunneling parameters at the next moment is determined based on the fusion characteristics.
[0026] In some examples, it also includes:
[0027] Based on the difference between the final prediction result and the actual detection results of the tunnel boring machine's tunneling parameters, the predictive performance of the tunnel boring machine's tunneling parameter prediction method is evaluated to determine whether it meets the standard.
[0028] In some examples, the method further includes: when a failure is determined, fine-tuning the long short-term memory network model based on the loss between the output of the long short-term memory network model and the actual detection results of the tunnel boring machine's tunneling parameters.
[0029] In some examples, after obtaining the final prediction of the tunneling parameters at the next time step, the process also includes:
[0030] Based on the final prediction results, the main control parameters of the tunnel boring machine are adjusted.
[0031] Secondly, embodiments of this application provide a tunnel boring machine tunneling parameter prediction system, including:
[0032] The filtering module is used to sort the features by importance according to the tunneling parameters of the tunnel boring machine, and to filter out a refined feature set based on the sorting results.
[0033] The initial prediction result acquisition module is used to input the refined feature set into a pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step, and to perform wavelet transform on the refined feature set to extract local change features from the refined feature set, and to input the global trend and local change features into a pre-trained convolutional neural network model to obtain the second prediction result of the tunneling parameters at the next time step.
[0034] The final prediction result acquisition module is used to fuse the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment.
[0035] Thirdly, embodiments of this application provide a computing device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the tunneling parameter prediction method for the tunnel boring machine according to the first aspect and any possible method in the first aspect.
[0036] In this embodiment, the importance of features is first ranked according to the tunneling parameters of the tunnel boring machine, and a refined feature set is selected based on the ranking results. Then, the refined feature set is input into a pre-trained long short-term memory network model to obtain a first prediction result of the tunneling parameters at the next time step. Wavelet transform is then performed on the refined feature set to extract local change features. The global trend and local change features are then input into a pre-trained convolutional neural network model to obtain a second prediction result of the tunneling parameters at the next time step. Finally, the first and second prediction results are fused to obtain the final prediction result of the tunneling parameters at the next time step. This method effectively captures sudden local changes and overall trend evolution of the tunnel boring machine's operating parameters, improving prediction accuracy and model robustness. Consequently, the tunneling process can be optimized, improving construction efficiency and safety.
[0037] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description
[0038] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:
[0039] Figure 1 A flowchart of a tunnel boring machine tunneling parameter prediction method according to an embodiment of this application;
[0040] Figure 2 This is a schematic diagram illustrating the implementation process of a tunnel boring machine tunneling parameter prediction method according to an embodiment of this application;
[0041] Figure 3 This is a structural block diagram of a tunnel boring machine tunneling parameter prediction system according to an embodiment of this application;
[0042] Figure 4 A schematic diagram of a computing device suitable for implementing embodiments of this application is shown. Detailed Implementation
[0043] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the relevant application and not intended to limit the application. Furthermore, it should be noted that, for ease of description, only the parts relevant to the application are shown in the accompanying drawings.
[0044] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0045] The following describes, with reference to the accompanying drawings, a method, system, and device for predicting tunneling parameters of a tunnel boring machine according to embodiments of this application.
[0046] The tunnel boring machine (TBM) tunneling parameter prediction method, system and equipment of this application, by introducing neural networks and real-time monitoring of the current tunneling parameters of the TBM, can continuously and dynamically predict the tunneling parameters (which can also be called main control parameters or operating parameters) during the TBM tunneling process. Of course, in some examples, the tunneling parameters of the TBM can also be dynamically adjusted according to the prediction results.
[0047] Figure 1 This is a flowchart of a tunnel boring machine tunneling parameter prediction method according to an embodiment of this application, as shown below. Figure 1 As shown, a method for predicting tunneling parameters of a tunnel boring machine according to an embodiment of this application includes the following steps:
[0048] S101: Sort the features by importance according to the tunneling parameters of the tunnel boring machine, and select a refined feature set based on the sorting results.
[0049] The tunneling parameters of a tunnel boring machine include, but are not limited to: tunneling speed, cutterhead torque, total thrust, penetration depth, cutterhead speed, motor current, and motor voltage.
[0050] First, the current tunneling parameters of the tunnel boring machine (TBM) are collected, for example, continuously collecting the tunneling parameters for each tunneling cycle during normal tunneling. Then, the tunneling parameters of the TBM are sorted by feature importance, and a refined feature set is selected based on the sorting results.
[0051] As a specific example, the process involves ranking features based on their importance according to the tunneling parameters of a tunnel boring machine (TBM), and then selecting a refined feature set based on the ranking results. This includes: obtaining the TBM's tunneling parameters; standardizing the TBM's tunneling parameters to obtain standardized features for each parameter; evaluating the information gain of the standardized features for each parameter across all tree species; ranking the standardized features of each parameter based on the information gain; and obtaining the refined feature set based on the standardized features of the top N most important tunneling parameters, where N is less than the number of types of tunneling parameters.
[0052] Further, a refined feature set is selected based on the sorting results, specifically including: constructing time series data in the form of a sliding window for each tunneling parameter corresponding to the refined feature set in chronological order; and obtaining the refined feature set based on the time series data.
[0053] In practical applications, the XGBoost network model can be used to filter the tunneling parameters of a tunnel boring machine, thereby selecting a refined feature set. For example, the input of the XGBoost network model is the tunneling parameters of the tunnel boring machine, and the output is the feature importance ranking. The top 20 features in the feature importance ranking are selected to form a refined feature set.
[0054] Combination Figure 2 As shown, dynamic operating parameter data (i.e., tunneling parameters) of a tunnel boring machine (TBM) are continuously collected during each tunneling cycle in the normal tunneling process. Then, all valid tunneling parameters collected by the TBM except for the prediction target parameters are used as input. The XGBoost network model is used to sort the features by importance, and the top N features, such as the top 20 features, are extracted. Their contribution to the prediction performance is quantified and sorted to form a refined feature set.
[0055] Specifically, the Gain metric is used to measure feature importance and generate a ranking of feature importance, including:
[0056] The tunneling parameters of the tunnel boring machine are standardized to ensure that the mean of the features is 0 and the variance is 1. The standardization formula is shown in Formula 1:
[0057]
[0058] Where X is the original feature, μ is the mean of the feature, and σ is the standard deviation of the feature. scaled These are the standardized features.
[0059] The Gain metric is used to evaluate the information gain of a feature across all tree types. Let the loss function of the i-th feature at the k-th split point in j trees be Gain. ijk The total importance of this feature is shown in Formula 2:
[0060]
[0061] Where, δ ijk If feature i is used in the k-th node of tree j, then the value is 1.
[0062] Arrange the importance from highest to lowest, and the calculation formula is shown in Formula 3:
[0063] SortedFeatures=argsort(Importance)↓ (3)
[0064] Finally, the select_top_features(model, top_k) function is used to extract the top 20 features by feature importance.
[0065] The tunneling parameter data corresponding to the top 20 features in terms of feature importance are constructed into a time series data in the form of a sliding window, denoted as X. ts A window is created every 5 seconds with a sliding interval of 1 second, thus obtaining a refined feature set.
[0066] S102: Input the refined feature set into the pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step.
[0067] That is, the refined feature set is used as the input of the Long Short-Term Memory (LSTM) network model. The LSTM network model is pre-trained, so the prediction result of the next time step of the tunneling parameters can be output by the LSTM network model. This prediction result becomes the first prediction result.
[0068] S103: Perform wavelet transform on the refined feature set to extract local variation features from the refined feature set.
[0069] In one embodiment of this application, wavelet transform is performed on the refined feature set to extract local variation features from the refined feature set, including: using multi-scale wavelet decomposition to perform wavelet transform operation on the features in the refined feature set to obtain approximation coefficients and detail coefficients at multiple scales; forming a wavelet feature matrix based on the approximation coefficients and detail coefficients at the multiple scales, wherein the wavelet feature matrix represents the local variation features.
[0070] Combination Figure 2 As shown, the features processed by wavelet transform are used as input to the convolutional neural network model (CNN) to enhance the model's ability to perceive changes at different scales.
[0071] Specifically, for the local variation features at different frequencies, multi-scale wavelet decomposition (Daubechies wavelet, level=2) is used to transform each input variable. Wavelet decomposition yields approximation coefficients and detail coefficients at multiple scales, which are then concatenated to form the wavelet feature matrix X. pm , which serves as the input to the convolutional neural network model CNN.
[0072] S104: Input the global trend and local change features into the pre-trained convolutional neural network model to obtain the second prediction result of the tunneling parameters at the next time step.
[0073] S105: Combine the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment.
[0074] The method of fusing the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment includes: splicing the first prediction result and the second prediction result together, and obtaining a weighting coefficient based on the splicing result; performing a weighted fusion of the first prediction result and the second prediction result based on the weighting coefficient to obtain a fusion feature; and determining the final prediction result of the tunneling parameters at the next moment based on the fusion feature.
[0075] Then, based on the difference between the final prediction result and the actual detection result of the tunnel boring machine's tunneling parameters, the prediction performance of the tunnel boring machine's tunneling parameter prediction method is evaluated to determine whether it meets the standard.
[0076] When it is determined that the target is not met, the Long Short-Term Memory Network Model is fine-tuned based on the loss between the output of the Long Short-Term Memory Network Model and the actual detection results of the tunneling parameters of the tunnel boring machine.
[0077] After obtaining the final prediction result of the tunneling parameters at the next moment, the method further includes: adjusting the main control parameters of the tunnel boring machine based on the final prediction result.
[0078] As a concrete example, combined Figure 2 As shown, the data after wavelet transform is divided into two sets: 80% for training and 20% for testing. A CNN-LSTM model is constructed and trained, and an attention mechanism is introduced to concatenate the CNN and LSTM outputs to construct a joint feature representation, thereby achieving dynamic output of the TBM's tunneling parameter prediction results.
[0079] The prediction results of the mixture model are evaluated using test set data. The evaluation metric is the coefficient of determination (R²). 2 ), mean absolute error (MAE), and root square error (RMSE).
[0080] X pm The input and output of the CNN model pass through three one-dimensional convolutional layers (Conv1d) with a kernel size of 1 and a channel count of [32, 32, 32]. At the end, global average pooling is used to compress the three-dimensional feature map into a fixed-length one-dimensional vector h. CNN .
[0081] After each convolutional operation, the CNN neural network model uses the ReLU activation function to enhance its nonlinear expression ability, helping the model capture and express more complex feature changes and improve its sensitivity to staged or abrupt changes during the TBM tunneling process.
[0082] Standardized X tsThe input is an LSTM model, and the second neural network model uses the LSTM network structure, including one LSTM hidden layer with 64 neurons, and outputs the hidden state h at the last time step. LSTM A fully connected layer is connected after the hidden layer.
[0083] The LSTM neural network model is trained using the Adam optimizer with an initial learning rate of 0.0001 and a loss function of mean squared error, which measures the difference between the model's predicted values and the actual values. During training, the data is divided into training and validation sets, and the model weights and bias parameters are optimized through iterative iteration.
[0084] The vector h output by the CNN module CNN With the vector h output by the LSTM LSTM The concatenated values are then fed into a fully connected attention layer to calculate the attention weight coefficients. The attention scores, after softmax normalization, are used to perform weighted fusion of the features, resulting in a joint feature vector, as shown in Equation 5.
[0085] α = softmax(W) α [h LSTM ||h CNN ]),
[0086] H fused =α·[h LSTM ||h CNN (5)
[0087] Fusion feature H fused The predicted results of the target tunneling parameters are mapped to the fully connected layer. The model training uses mean squared error (MSE) as the main loss function and introduces L2 regularization to suppress overfitting, as shown in Equation 6:
[0088]
[0089] The model's predictions are compared and evaluated with actual monitoring results. The prediction results of the hybrid model are assessed using test set data. The evaluation metric is the coefficient of determination (R²). 2 Mean Absolute Error (MAE) and Root Squared Error (RMSE), for example: MAE, RMSE, and R 2 As shown in Formula 7:
[0090]
[0091] Among them, y i Let i be the i-th true value in the test set. is the predicted value of the mixture model, and n is the number of samples in the test set.
[0092] During training, the Adam optimizer was used, combined with the ReduceLROnPlateau strategy to automatically reduce the learning rate when there was no improvement in validation set performance. Meanwhile, R... 2 The model with the highest performance and lowest MRE is used as the final prediction model. If the model performance does not meet the target (e.g., R...), then... 2 If the regularization strength λ is less than 0.6 and MRE is greater than 0.4, then the regularization strength λ will be automatically increased. reg This is to enhance the model's generalization ability.
[0093] The tunnel boring machine (TBM) parameter prediction method of this application, by introducing neural networks and real-time data monitoring, can continuously and dynamically predict the main control parameters during the TBM tunneling process, and dynamically adjust the tunneling setting parameters (i.e., main control parameters) according to the prediction results.
[0094] The embodiments of this application design two neural network models. The first XGBoost network model takes the TBM's operating parameters as input and outputs a feature importance ranking, selecting the top 20 features to form a refined feature set. The second neural network model uses the refined feature set output by the first XGBoost model, extracts multi-scale features from the time series through wavelet transform, uses the original refined time series features as input to an LSTM, and uses the refined features after wavelet transform as input to a CNN. An attention mechanism is introduced, and the CNN output and LSTM output are concatenated to predict the TBM's operating parameters at the next time step, thereby achieving dynamic prediction of key parameters during TBM construction.
[0095] A large amount of historical tunneling data was used to train and optimize the neural network model. In the feature selection stage, an XGBoost model was introduced to analyze the feature importance of tunneling parameters. This effectively captures the complex nonlinear relationships between TBM operating parameters, identifies key feature variables that significantly affect target prediction parameters, and constructs a refined input feature set, improving the model's training efficiency and prediction accuracy. In the time-series modeling stage, wavelet transform was used to decompose the time-series data at multiple scales, extracting local variation features at different frequencies, enhancing the model's ability to express non-stationary features such as signal abrupt changes and phased changes. The original time-series data was used as input to the LSTM network to model long-term dependencies and evolutionary trends; the wavelet-transformed features were used as input to the CNN network to extract short-term dynamic changes and local spatial features. A CNN-LSTM fusion model was introduced, fully combining the advantages of convolutional neural networks (CNN) in capturing local patterns and wavelet details with the capabilities of long short-term memory networks (LSTM) in handling time dependencies and trend modeling, achieving more comprehensive and refined dynamic prediction of TBM tunneling parameters. Meanwhile, an attention mechanism is introduced to weight and fuse the outputs of CNN and LSTM, dynamically adjusting the importance of different features. This enables the model to adaptively focus on key features according to different scenarios, effectively improving the model's performance in prediction tasks under complex conditions.
[0096] The model employs mean squared error (MSE) as the primary loss function to measure the difference between the model's predicted and actual values. An L2 regularization term is introduced to suppress model complexity and prevent overfitting. During training, an early stopping mechanism and an adaptive learning rate adjustment strategy (such as ReduceLROnPlateau) are combined to automatically stop training or reduce the learning rate when the model's performance on the validation set shows no improvement, thereby enhancing the model's generalization ability and robustness.
[0097] During actual construction, the TBM's operating parameters are collected in real time and input into a neural network model, outputting real-time predictions of key parameters. Based on the prediction results, tunneling parameters such as tunneling speed, cutterhead torque, and total thrust are dynamically adjusted to optimize the tunneling scheme and ensure the safety and efficiency of TBM tunneling under complex geological conditions. This system is particularly suitable for TBM construction under complex geological conditions. Through the real-time dynamic prediction system, it can effectively capture sudden local changes and overall trend evolution of TBM operating parameters, thereby improving prediction accuracy and model robustness.
[0098] The tunnel boring machine (TBM) parameter prediction method according to embodiments of this application first ranks the TBM's tunneling parameters by feature importance and then selects a refined feature set based on the ranking results. Next, the refined feature set is input into a pre-trained long short-term memory (LSTM) network model to obtain a first prediction result for the tunneling parameters at the next time step. Wavelet transform is then applied to the refined feature set to extract local change features. The global trend and local change features are then input into a pre-trained convolutional neural network model to obtain a second prediction result for the tunneling parameters at the next time step. Finally, the first and second prediction results are fused to obtain the final prediction result for the tunneling parameters at the next time step. This method effectively captures sudden local changes and overall trend evolution of TBM operating parameters, improving prediction accuracy and model robustness. Consequently, it can optimize the tunneling process and improve construction efficiency and safety.
[0099] Figure 3 This is a structural block diagram of a tunnel boring machine tunneling parameter prediction system according to an embodiment of this application, as shown below. Figure 3 As shown, a tunnel boring machine tunneling parameter prediction system according to an embodiment of this application includes: a screening module 310, an initial prediction result acquisition module 320, and a final prediction result acquisition module 330, wherein:
[0100] The filtering module 310 is used to sort the features by importance according to the tunneling parameters of the tunnel boring machine, and to filter out a refined feature set based on the sorting results.
[0101] The initial prediction result acquisition module 320 is used to input the refined feature set into a pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step, and to perform wavelet transform on the refined feature set to extract local change features from the refined feature set, and to input the global trend and local change features into a pre-trained convolutional neural network model to obtain the second prediction result of the tunneling parameters at the next time step.
[0102] The final prediction result acquisition module 330 is used to fuse the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment.
[0103] The tunnel boring machine (TBM) parameter prediction system according to an embodiment of this application first ranks the TBM's tunneling parameters by feature importance and then selects a refined feature set based on the ranking results. Next, the refined feature set is input into a pre-trained long short-term memory (LSTM) network model to obtain a first prediction result for the tunneling parameters at the next time step. Wavelet transform is then applied to the refined feature set to extract local change features. The global trend and local change features are then input into a pre-trained convolutional neural network model to obtain a second prediction result for the tunneling parameters at the next time step. Finally, the first and second prediction results are fused to obtain the final prediction result for the tunneling parameters at the next time step. This system effectively captures sudden local changes and overall trend evolution of TBM operating parameters, improving prediction accuracy and model robustness. Consequently, it can optimize the tunneling process and improve construction efficiency and safety.
[0104] It should be noted that the specific implementation of the tunnel boring machine tunneling parameter prediction system in this application embodiment is similar to the specific implementation of the tunnel boring machine tunneling parameter prediction method in this application embodiment. Please refer to the description in the method section for details, which will not be repeated here.
[0105] The following is for reference. Figure 4 , Figure 4 A schematic diagram of a computing device structure suitable for implementing embodiments of this application is shown.
[0106] like Figure 4 As shown, the computer system includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 1002 or programs loaded from storage section 1008 into random access memory (RAM) 1003. RAM 1003 also stores various programs and data required for the system's operating instructions. CPU 1001, ROM 1002, and RAM 1003 are interconnected via bus 1004. Input / output (I / O) interface 1005 is also connected to bus 1004.
[0107] The following components are connected to I / O interface 1005: an input section 1006 including a keyboard, mouse, etc.; an output section 1007 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to I / O interface 1005 as needed. A removable medium 1011, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 1010 as needed so that computer programs read from it can be installed into storage section 1008 as needed.
[0108] Specifically, according to embodiments of this application, the flowchart above refers to... Figure 1 The described process can be implemented as a computer software program. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such an embodiment, the computer program contains program code for performing the methods shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via communication section 1009, and / or installed from removable medium 1011. When the computer program is executed by central processing unit (CPU) 1001, it performs the functions defined in the system of this application.
[0109] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0110] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operational instructions of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two connected blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified functions or operational instructions, or using a combination of dedicated hardware and computer instructions.
[0111] The units or modules described in the embodiments of this application can be implemented in software or hardware. The described units or modules can also be located in a processor. The names of these units or modules do not, in certain circumstances, constitute a limitation on the unit or module itself.
[0112] In another aspect, this application also provides a computer-readable storage medium, which may be included in the computing device described in the above embodiments, or may exist independently and not assembled into the computing device. The computer-readable storage medium stores one or more programs that, when used by one or more processors, execute the tunnel boring machine parameter prediction method described in this application. Specifically, it ranks features based on their importance according to the tunnel boring machine's parameters and selects a refined feature set based on the ranking results.
[0113] The refined feature set is input into a pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step.
[0114] Wavelet transform is applied to the refined feature set to extract local variation features from the refined feature set;
[0115] The global trend and local change features are input into a pre-trained convolutional neural network model to obtain the second prediction result of the tunneling parameters at the next time step.
[0116] By combining the first prediction result and the second prediction result, the final prediction result of the tunneling parameters at the next moment is obtained.
[0117] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.
Claims
1. A method for predicting tunneling parameters of a tunnel boring machine, characterized in that, include: The importance of features is ranked according to the tunneling parameters of the tunnel boring machine, and a refined feature set is selected based on the ranking results. The refined feature set is input into a pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step. Wavelet transform is applied to the refined feature set to extract local variation features from the refined feature set; By inputting global trend and local change features into a pre-trained convolutional neural network model, a second prediction result of the tunneling parameters at the next time step is obtained; By combining the first prediction result and the second prediction result, the final prediction result of the tunneling parameters at the next moment is obtained.
2. The method for predicting tunneling parameters of a tunnel boring machine according to claim 1, characterized in that, The process involves ranking features based on their importance according to the tunneling parameters of the tunnel boring machine, and then selecting a refined feature set based on the ranking results, including: Obtain the tunneling parameters of the tunnel boring machine; The tunneling parameters of the tunnel boring machine are standardized to obtain the standardized characteristics of each tunneling parameter; Evaluate the information gain of the standardized features of each tunneling parameter across all tree species; The importance of the standardized features of each tunneling parameter is ranked according to the information gain. The refined feature set is obtained based on the standardized features of the N most important tunneling parameters, where N is less than the number of types of tunneling parameters.
3. The method for predicting tunneling parameters of a tunnel boring machine according to claim 1 or 2, characterized in that, The process of selecting a refined feature set based on the sorting results specifically includes: The tunneling parameters corresponding to the refined feature set are constructed as time-series data in the form of a sliding window in chronological order; The refined feature set is obtained based on the time series data.
4. The method for predicting tunneling parameters of a tunnel boring machine according to claim 3, characterized in that, The step of performing wavelet transform on the refined feature set to extract local variation features from the refined feature set includes: Multi-scale wavelet decomposition is used to perform wavelet transform operations on the features in the refined feature set to obtain approximation coefficients and detail coefficients at multiple scales. Based on the approximation coefficients and detail coefficients at the multiple scales, a wavelet feature matrix is formed, wherein the wavelet feature matrix characterizes the local variation features.
5. The method for predicting tunneling parameters of a tunnel boring machine according to claim 1, characterized in that, The process of fusing the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment includes: The first prediction result and the second prediction result are concatenated, and the weighting coefficients are obtained based on the concatenation result; Based on the weighting coefficients, the first prediction result and the second prediction result are weighted and fused to obtain the fused feature; The final prediction result of the tunneling parameters at the next moment is determined based on the fusion characteristics.
6. The method for predicting tunneling parameters of a tunnel boring machine according to claim 1, characterized in that, Also includes: Based on the difference between the final prediction result and the actual detection results of the tunnel boring machine's tunneling parameters, the predictive performance of the tunnel boring machine's tunneling parameter prediction method is evaluated to determine whether it meets the standard.
7. The method for predicting tunneling parameters of a tunnel boring machine according to claim 6, characterized in that, Also includes: When it is determined that the target is not met, the Long Short-Term Memory Network Model is fine-tuned based on the loss between the output of the Long Short-Term Memory Network Model and the actual detection results of the tunneling parameters of the tunnel boring machine.
8. The method for predicting tunneling parameters of a tunnel boring machine according to claim 1, characterized in that, After obtaining the final prediction results of the tunneling parameters for the next time step, the following is also included: Based on the final prediction results, the main control parameters of the tunnel boring machine are adjusted.
9. A tunnel boring machine tunneling parameter prediction system, characterized in that, include: The filtering module is used to sort the features by importance according to the tunneling parameters of the tunnel boring machine, and to filter out a refined feature set based on the sorting results. The initial prediction result acquisition module is used to input the refined feature set into a pre-trained long short-term memory network model to obtain the first prediction result of the tunneling parameters at the next time step, and to perform wavelet transform on the refined feature set to extract local change features from the refined feature set, and to input the global trend and local change features into a pre-trained convolutional neural network model to obtain the second prediction result of the tunneling parameters at the next time step. The final prediction result acquisition module is used to fuse the first prediction result and the second prediction result to obtain the final prediction result of the tunneling parameters at the next moment.
10. A computing device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, it implements a method for predicting tunneling parameters of a tunnel boring machine according to any one of claims 1-8.