Deep hard rock tunnel TBM rapid tunneling method and device

Abstract

The invention relates to the technical field of tunneling, in particular to a deep hard rock tunnel TBM rapid tunneling method and device.The method comprises the following steps that S1, a horizontal directional drilling machine is used for conducting advanced drilling, and the torque of a drill bit is monitored in real time; s2, a machine learning and deep learning algorithm combining a convolutional neural network and a support vector machine is adopted to carry out stratum image recognition on rock core samples transported out through advanced drilling; s3, based on drill bit torque monitoring and stratum image recognition results, precise explosive distribution and blasting operation are implemented; and S4, TBM tunneling operation is carried out. The problems that in the prior art, deep-buried hard rock tunnel tunneling efficiency is low, cutter abrasion frequency is high, and operation danger is high are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel excavation, and more particularly to a method and device for rapid excavation of a deep-buried hard rock tunnel using a TBM. Background Art

[0002] With the rapid development of global infrastructure construction, tunnel boring machines (TBMs) have become a staple of tunnel construction, particularly in the construction of long and deep tunnels. TBMs have been widely used due to their full-face excavation, efficient construction, and excellent safety. However, when working in complex hard rock formations, TBM construction still faces numerous challenges, the most prominent of which is high tool wear. This problem not only seriously affects construction efficiency but also poses significant safety risks to on-site workers, limiting the effectiveness of TBMs in complex geological conditions.

[0003] First, in hard rock formations, due to the rock's extreme strength and hardness, TBM cutters are subjected to tremendous friction and impact forces, resulting in a much higher rate of cutter wear than in soft rock or medium-hard formations. During excavation, the cutter life is significantly shortened, forcing construction teams to frequently change cutters, which not only increases tool consumption costs but also disrupts construction progress. Frequent cutter changes significantly reduce the TBM's overall excavation efficiency and extend the construction period. Each cutter change requires the TBM to stop operating, forcing the entire construction process to a standstill and significantly reducing production efficiency.

[0004] Furthermore, frequent tool changes pose significant risks in hard rock formations, especially during deep tunnel construction. Due to the concentrated stress in the surrounding rock of hard rock formations, construction workers need to enter the tunnel face to perform tool changes. This area is often prone to rockbursts. Rockbursts typically occur when surrounding rock stress redistributes, causing the rock formation to suddenly fracture, ejecting large amounts of rock blocks and posing a significant risk to the lives of construction workers. Rockbursts can not only directly injure construction workers but can also damage the tunnel's support structures, leading to further construction delays and safety accidents. Especially in deep hard rock tunnels, the high stress and complex geological conditions significantly increase the probability of rockbursts during tool changes. Frequent tool changes significantly increase the time construction workers are exposed to the risk of rockbursts, threatening their lives.

[0005] To address these issues, prior art typically employs grouting to reinforce the surrounding rock near the tunnel face during TBM cutter changes. Grouting strengthens the surrounding rock by injecting slurry into the surrounding rock, increasing its overall strength and facilitating cutter opening and cutter changes. However, this traditional method has numerous drawbacks. First, grouting is extremely expensive, particularly in deep tunnels and hard rock formations, where the materials and equipment required for grouting are expensive, leading to high construction costs. Second, grouting takes a long time to implement, requiring interruptions during construction, significantly delaying the project. Grouting not only requires significant construction time, but also poses challenges in ensuring accurate injection of the slurry into the areas requiring reinforcement, making precise grouting placement impossible. The random nature of slurry diffusion often results in suboptimal reinforcement, increasing uncertainty during subsequent construction. More importantly, while grouting can strengthen the rock near the tunnel face and reduce the risk of workers having to open the cutter for cutter changes, it does not fundamentally address the issues of rapid cutter wear, frequent cutter changes, and low tunneling efficiency. Summary of the Invention

[0006] The present invention aims to overcome at least one defect (shortcoming) of the above-mentioned prior art and provide a method for rapid excavation of deep hard rock tunnels by TBM, which is used to solve the problems of low excavation efficiency, high frequency of tool wear and high operation risk in the prior art.

[0007] The technical solution adopted by the present invention is a method for rapid excavation of a deep hard rock tunnel by a TBM, the method comprising the following steps:

[0008] S1: Use a horizontal directional drill to conduct advance drilling and monitor the torque of the drill bit in real time;

[0009] S2: Uses a machine learning and deep learning algorithm combining convolutional neural networks and support vector machines to perform stratigraphic image recognition on core samples transported from advance drilling.

[0010] S3: Implement precise explosive placement and blasting operations based on drill bit torque monitoring and formation image recognition results;

[0011] S4: Perform TBM excavation operations.

[0012] When constructing some long and large tunnels and deep buried tunnels, in order to avoid affecting the construction efficiency and posing huge safety hazards to the workers, in this application, advance drilling is first carried out, and then the core samples transported out by drilling are subjected to stratum image recognition. Then, based on the identified stratum results and the drill bit torque monitored during the advance drilling, accurate explosives are placed according to the current geological structure of the tunnel, and then precise blasting is carried out to release the surrounding rock pressure and loosen the rock layer near the tunnel axis, creating an environment conducive to TBM excavation. After the blasting is completed and the excavation area is confirmed to be safe, the TBM excavation operation is carried out, thereby avoiding frequent tool replacement, extending the service life of the tool, reducing the number of times the construction process is interrupted, improving production operation efficiency, reducing construction costs without grouting reinforcement, and ensuring the safety of the construction process and the safety of life and property of construction workers.

[0013] Preferably, step S2 includes the following steps:

[0014] S21: Capture images of core samples transported from advance drilling and perform data preprocessing on the images;

[0015] S22: Using convolutional neural network to extract features from preprocessed image data;

[0016] S23: Input the extracted features into the support vector machine for classification;

[0017] S24: Compare the classification results with the stratigraphic database to identify different stratigraphic types and generate a geological report.

[0018] In this application, the stratigraphic images are processed and analyzed by combining convolutional neural networks and support vector machines, which fully utilizes the advantages of both to improve the performance of image classification and feature recognition, so that complex stratigraphic structures can be accurately identified and classified. Secondly, a geological report is generated based on the identification and classification results, which makes it easier for staff to understand the geological conditions and carry out explosives deployment and blasting operations.

[0019] Preferably, the step S21 includes: performing image enhancement using data enhancement technology and performing image normalization processing.

[0020] Further preferably, the normalization processing formula is:

[0021]

[0022] Among them, I raw is the original image pixel value, min(I) and max(I) are the minimum and maximum values ​​of the image pixels respectively, I norm is the normalized image pixel value.

[0023] Through data augmentation technology, the collected data is rotated, flipped, cropped, scaled, and processed to create new data samples, so that the model can be exposed to a wider variety of samples during training, better learn features and improve robustness. Then, the data after image enhancement is normalized to normalize the image pixel values ​​to the range of [0,1], so that the scale of the input data will be more consistent, avoiding the problem of gradient disappearance or explosion due to large differences in data scales, which helps to improve training efficiency, make the model more likely to converge, and improve model performance.

[0024] Preferably, step S22 includes extracting local features of the image through a convolution operation, and the formula of the convolution operation is:

[0025]

[0026] Where f(x,y) represents the convolution result, I(x,y) is the original pixel value of the image, K(i,j) is the convolution kernel, k is the size of the convolution kernel, and x and y are the image coordinates.

[0027] Stratigraphic images usually have complex textures and structures. Convolutional neural networks (CNNs) can automatically extract these complex spatial features and provide better recognition capabilities in the diversity and complexity of stratigraphic images. Due to environmental changes and different shooting angles, stratigraphic images may be subject to certain noise interference. The features extracted by convolutional neural networks through operations such as convolution and pooling have good robustness and accuracy, thereby improving the accuracy of geological image recognition.

[0028] Preferably, in step S23, a classification rule is formulated through a classification decision function, and then a suitable kernel function is selected to distinguish different stratum types; wherein the classification decision function formula is:

[0029]

[0030] Among them, x is the sample to be classified, x i is the support vector, α i is the Lagrange multiplier, y i is the sample label, K(x i ,x) is the kernel function, b is the bias term, and sign is the sign function.

[0031] Preferably, the kernel function is a radial basis function, which is used to capture complex features in the image and thus distinguish different stratum types; wherein the formula of the radial basis function is:

[0032]

[0033] Among them, σ represents a hyperparameter; |x i -x j | represents sample x i and x j The Euclidean distance between .

[0034] After using CNN for feature extraction, support vector machine (SVM) is also used for classification. The classification boundary of SVM can ensure the robustness and accuracy of classification and efficiently identify complex image features. At the same time, SVM can also maintain good generalization ability in the feature space, reducing the risk of overfitting, thereby ensuring the accuracy of geological image recognition.

[0035] Preferably, the step S24 includes:

[0036] S241: Designing a stratum database according to information of different stratum types stored in the database;

[0037] S242: Initiating a query to the stratum database based on the classification result label and the feature extraction result, and retrieving the stratum database to obtain stratum information related to the classification result;

[0038] S243: The retrieved stratigraphic information is compared with the stratigraphic image recognition result. If they are consistent, the classification result of the stratigraphic image is correct. If they are inconsistent, the feedback mechanism is triggered to perform warning and feedback operations, trigger manual inspection or further analysis, and finally generate a final geological report.

[0039] In this application, a detailed geological report is generated to provide an important basis for the specific location of subsequent explosives deployment, ensuring that blasting operations can be carried out accurately in the target area, thereby improving the safety of construction.

[0040] Preferably, step S4 also includes real-time monitoring of the operating status of the TBM, collecting the working data of the TBM in real time through sensors, and establishing a visual monitoring platform to display all data in real time. At the same time, the excavation parameters can be adjusted according to the monitored excavation operation conditions.

[0041] During excavation operations, the TBM's operating status is also monitored to ensure that the TBM advances steadily along the tunnel axis and avoids deviation in any direction. The excavation speed can also be adjusted at any time based on monitoring data to ensure the efficient operation of the TBM. At the same time, the tool can be replaced and maintained based on real-time monitoring data to prevent excavation delays or safety risks caused by excessive tool wear.

[0042] On the other hand, the present application also provides a deep hard rock tunnel TBM rapid excavation device based on the above-mentioned deep hard rock tunnel TBM rapid excavation method, the device comprising:

[0043] Drilling module: used for advanced drilling using a horizontal directional drill and real-time monitoring of the drill bit torque;

[0044] Image recognition module: This module uses a machine learning and deep learning algorithm that combines a convolutional neural network with a support vector machine to perform formation image recognition on core samples transported from advance drilling.

[0045] Blasting module: used to implement precise explosive placement and blasting operations based on drill bit torque monitoring and formation image recognition results;

[0046] Tunneling operation module: used for TBM tunneling operations.

[0047] This system combines advanced drilling, image recognition technology and precision blasting technology, which can effectively reduce the wear frequency of TBM cutters, improve excavation efficiency and reduce the risk of rock burst.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] This invention can effectively improve the excavation efficiency of deep hard rock tunnels, reduce tool wear and the number of tool changes, lower the cost of grouting reinforcement during tool opening and tool change, and reduce the risks associated with tool change operations for workers, thereby reducing safety risks and costs during construction. Furthermore, the technical solution of combining advance drilling with precision blasting can better adapt to complex geological conditions, preemptively release rock pressure in the area to be excavated, and ensure precise loosening of the rock formation near the tunnel axis, creating favorable conditions for rapid TBM excavation. This innovative method will significantly shorten the construction period, improve the safety and economic efficiency of tunnel excavation, and has promising application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Flow chart of the method of the present invention.

[0051] Figure 2 This is the overall structural diagram of the TBM rapid excavation method of the present invention.

[0052] Figure 3 This is a flow chart of the formation image recognition of the present invention.

[0053] Figure 4 It is a schematic diagram of the blasting process of the present invention. DETAILED DESCRIPTION

[0054] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting the present invention. To better illustrate the following embodiments, some components in the accompanying drawings may be omitted, enlarged, or reduced in size, and do not represent actual product dimensions. Those skilled in the art will appreciate that some well-known structures and their descriptions may be omitted from the accompanying drawings.

[0055] Example 1

[0056] like Figure 1 As shown, this embodiment provides a method for rapid excavation of a deep hard rock tunnel by a TBM, the method comprising the following steps:

[0057] Step S1: using a horizontal directional drill to perform advance drilling and monitor the torque of the drill bit in real time;

[0058] like Figure 2 As shown, the step S1 includes:

[0059] Step S11: Use a horizontal directional drill (HDD) to drill from the surface. After reaching the depth of the tunnel axis, ensure that the drill bit advances along the tunnel axis. During drilling, use a torque monitoring device to monitor the drill bit's torque changes in real time to determine the hardness of the rock formation. When the rock formation is hard, the drill bit's torque will be high, while when the rock formation is soft, the drill bit's torque will be relatively low. The drill bit's torque changes provide an auxiliary basis for analyzing the rock formation hardness, providing information for subsequent explosives placement and precision blasting. After the drill bit passes through the blasting area along the tunnel axis, it drills upward to the ground at a suitable location for subsequent explosives placement. In addition, operators need to regularly maintain and calibrate the HDD to ensure that the equipment operates in optimal condition.

[0060] Step S12: Core samples extracted during the drilling process are properly stored and labeled. These samples are photographed using high-resolution image acquisition equipment to record their appearance and characteristics. To ensure the accuracy of subsequent stratigraphic analysis, core samples must remain intact and accurately labeled according to the depth of collection. This allows for accurate stratum localization during subsequent image recognition and geological analysis.

[0061] Step S13: Throughout the drilling process, the status of the drilling equipment, especially the progress of the drilling, is regularly checked to ensure the safety and stability of the drilling process. If any abnormalities are found, such as severe equipment wear or deviation of the drill hole, adjustments should be taken immediately to ensure the accuracy and continuity of the drilling.

[0062] Preferably, in step S2, a machine learning and deep learning algorithm combining a convolutional neural network and a support vector machine is used to perform formation image recognition on the core samples transported out by the advance drilling;

[0063] Specifically, if Figure 3 As shown, the step S2 includes:

[0064] Step S21: collecting images of the core samples transported by the advance drilling, and performing data preprocessing on the images;

[0065] Specifically, images of the core samples drilled out in advance are collected, and high-definition cameras and professional imaging equipment are used to comprehensively photograph the samples to obtain high-definition stratigraphic images, ensuring that the images obtained are clear and can display the geological characteristics of the samples, especially in hard rock formations. The texture, cracks and other characteristics of the rock are very critical for subsequent analysis.

[0066] Secondly, in order to improve the quality of the image and enhance the robustness of the model, data augmentation technology is used for image enhancement. Processing methods include rotation, flipping, cropping, and scale transformation. Image normalization is also performed. The following formula is used to normalize the image pixel values ​​to the range of [0, 1]:

[0067]

[0068] Among them, I raw is the original image pixel value, min(I) and max(I) are the minimum and maximum values ​​of the image pixels respectively, I norm is the normalized image pixel value.

[0069] Therefore, in this embodiment, the collected data is rotated, flipped, cropped, scaled, etc. through data enhancement technology to create new data samples, so that the model can be exposed to more types of samples during training, can better learn features and improve robustness. Then, the data after image enhancement is normalized to normalize the image pixel values ​​to the range of [0,1], so that the scale of the input data will be more consistent, avoiding the problem of gradient disappearance or explosion due to excessive differences in data scales, which helps to improve training efficiency, makes the model more likely to converge, and improves the performance of the model.

[0070] Step S22: using a convolutional neural network to perform feature extraction on the preprocessed image data;

[0071] Preferably, step S22 includes extracting local features of the image through a convolution operation, including features such as the texture, weathering, and color of the rock mass. By extracting the image features, the type of rock layer, the hardness of the rock, and the distribution of cracks can be identified. The formula for the convolution operation is:

[0072]

[0073] Where f(x,y) represents the convolution result, I(x,y) is the original pixel value of the image, K(i,j) is the convolution kernel, k is the size of the convolution kernel, and x and y are the image coordinates.

[0074] Stratigraphic images usually have complex textures and structures. Convolutional neural networks (CNNs) can automatically extract these complex spatial features and provide better recognition capabilities in the diversity and complexity of stratigraphic images. Due to environmental changes and different shooting angles, stratigraphic images may be subject to certain noise interference. The features extracted by convolutional neural networks through operations such as convolution and pooling have good robustness and accuracy, thereby improving the accuracy of geological image recognition.

[0075] Step S23: Input the extracted features into a support vector machine for classification. This method can accurately classify complex stratum structures and is more accurate than traditional methods in identifying stratum features.

[0076] Preferably, in step S23, a classification rule is formulated through a classification decision function, and then a suitable kernel function is selected to distinguish different stratum types; wherein the classification decision function formula is:

[0077]

[0078] Among them, x is the sample to be classified, x i is the support vector, α i is the Lagrange multiplier, y i is the sample label, K(x i ,x) is the kernel function, b is the bias term, and sign is the sign function. This decision function is used to classify different types of strata.

[0079] Preferably, due to the complex geological conditions, in the embodiment, a radial basis function (RBF kernel) is selected to distinguish different stratum types. The RBF kernel can effectively handle nonlinear relationships. It can map the input data to a high-dimensional space, so that the originally linearly inseparable situation becomes linearly separable. Therefore, for complex stratum image data, the RBF kernel can better capture the complex features in the image (such as the texture, color, structure, etc. of the rock layer). Moreover, the RBF kernel has good robustness to noise and local changes in samples. Taking into account that in actual tunnel engineering, the quality of stratum images may be affected by many factors (such as shooting angle, illumination changes, etc.), the RBF kernel can better cope with these changes. Therefore, in this embodiment, the radial basis function is used to capture the complex features in the image to distinguish different stratum types; wherein, the formula of the radial basis function is:

[0080]

[0081] Among them, σ is a hyperparameter that controls the influence range of data points in high-dimensional space. The optimal σ value can be selected through methods such as cross-validation and grid search; |x i -x j| represents sample x i and x j The Euclidean distance between them reflects their similarity in the input space.

[0082] Therefore, in this embodiment, after using CNN for feature extraction, support vector machine SVM is also used for classification. The classification boundary of SVM can ensure the robustness and accuracy of classification and efficiently identify complex image features. At the same time, SVM can also maintain good generalization ability in the feature space, reduce the risk of overfitting, and thus ensure the accuracy of geological image recognition.

[0083] Step S24: Compare the classification results with the stratigraphic database to confirm the specific characteristics of the current stratigraphic layer, identify different stratigraphic types and generate a geological report. The report not only includes relevant parameters such as the type and structure of the stratigraphic layer, but also provides an important basis for the specific location of subsequent explosives deployment, ensuring that the blasting operation can be carried out accurately in the target area.

[0084] Preferably, the step S24 includes:

[0085] Step S241: Designing a stratum database based on information on different stratum types stored in the database; the database may store information such as rock thickness, hardness, mineral composition, texture characteristics, and may even include acquired image data.

[0086] Step S242: After the formation image is classified by the CNN and SVM models, a classification label and its corresponding feature data are generated. Then, based on the classification result label and the feature extraction result, a query is initiated to the formation database, and formation information related to the classification result is retrieved from the formation database.

[0087] Step S243: The retrieved stratigraphic information is compared with the result of stratigraphic image recognition. If they are consistent, the classification result of the stratigraphic image is correct. If they are inconsistent, the feedback mechanism is triggered to perform warning and feedback operations, trigger manual inspection or further analysis, and finally generate a final geological report.

[0088] By generating a detailed geological report, an important basis is provided for the specific location of subsequent explosives deployment, ensuring that blasting operations can be carried out accurately in the target area, thereby improving construction safety.

[0089] Therefore, in this embodiment, the stratigraphic images are processed and analyzed by combining convolutional neural networks and support vector machines, which fully utilizes the advantages of both and improves the performance of image classification and feature recognition, so that complex stratigraphic structures can be accurately identified and classified. Secondly, a geological report is generated based on the identification and classification results, which makes it easier for staff to understand the geological conditions and carry out explosives deployment and blasting operations.

[0090] Preferably, the step S3 includes implementing accurate explosive placement and blasting operations based on drill bit torque monitoring and formation image recognition results;

[0091] Specifically, if Figure 2 and Figure 4 As shown, step S3 includes:

[0092] S31: Based on the drill bit torque information monitored in step S1 and the formation information identified in step S2, the location, depth, and quantity of explosives are determined. Especially in hard rock formations, precise placement of explosives is crucial to ensure that a sufficient area of ​​the rock is loosened during blasting without affecting the stability of other areas. In softer formations, a smaller amount of explosives, or even no explosives at all, may be deployed, depending on the specific circumstances.

[0093] S32: After determining the explosive placement plan, use a sufficiently long detonating cord to string all explosives together according to their specific locations. One end of the cord is attached to the drill bit at the advance drill exit. The drill bit is then pulled back to the start of the advance hole, placing the explosives at the designated location in the advance hole. The amount of explosives varies for different strata, and operators should adjust the amount based on drill bit torque data and information from the geological report to ensure economical use of explosives and improve tunneling efficiency.

[0094] S33: Before blasting, a comprehensive safety inspection must be conducted to ensure that all personnel and equipment have been evacuated to a safe area. Safety protection facilities in the blasting area should also be inspected and confirmed to be free of potential safety hazards before issuing the blasting signal, thereby ensuring the safety of the surrounding environment and workers.

[0095] S34: Carry out precision blasting, using pre-designed explosive placement plans to loosen the rock formations near the tunnel axis.

[0096] By evenly and accurately placing the explosives in the advance borehole and by properly arranging the explosives, it is possible to ensure that the shock wave generated during the blasting can effectively loosen the surrounding rock formations and reduce the rock strength, thereby facilitating subsequent TBM excavation and ensuring that the rock formations after blasting achieve the loosening effect required by the design.

[0097] Preferably, in step S4, after the blasting is completed, the TBM excavation operation is started. The loosened rock layer caused by the blasting will greatly reduce the excavation resistance of the TBM, reduce the wear on the tool, and improve the overall excavation efficiency.

[0098] Specifically, it also includes real-time monitoring of the TBM's operating status. During the excavation process, the TBM's working data is collected in real time through sensors such as tool wear, torque, vibration and thrust, and a visual monitoring platform is established to display all data in real time. The operator can intuitively view the TBM's operating status, monitor the rock hardness, tool status and equipment operation during the excavation process, especially the tool wear. If necessary, the tool can be replaced and maintained to prevent excavation delays or safety risks caused by excessive tool wear, and the excavation parameters can be adjusted as needed to ensure that the TBM advances steadily along the tunnel axis, avoid deviation in any direction, and ensure the safety and efficiency of construction.

[0099] This embodiment effectively improves the excavation efficiency of deep hard rock tunnels, reduces tool wear and tool changes, and reduces safety risks and costs during construction. Furthermore, the technical solution of combining advance drilling with precision blasting can better adapt to complex geological conditions, preemptively release rock pressure in the area to be excavated, and ensure precise loosening of the rock near the tunnel axis, creating favorable conditions for rapid TBM excavation. This innovative method will significantly shorten the construction period, improve the safety and economic efficiency of tunnel excavation, and has promising application prospects.

[0100] On the other hand, this embodiment further provides a deep hard rock tunnel TBM rapid excavation device based on the above-mentioned deep hard rock tunnel TBM rapid excavation method, the device comprising:

[0101] Drilling module: used for advanced drilling using a horizontal directional drill and real-time monitoring of the drill bit torque;

[0102] Image recognition module: This module uses a machine learning and deep learning algorithm that combines a convolutional neural network with a support vector machine to perform formation image recognition on core samples transported from advance drilling.

[0103] Blasting module: used to implement precise explosive placement and blasting operations based on drill bit torque monitoring and formation image recognition results;

[0104] Tunneling operation module: used for TBM tunneling operations.

[0105] The system described in this embodiment combines advanced drilling, image recognition technology and precision blasting technology, which can effectively reduce the wear frequency of TBM cutters, improve tunneling efficiency and reduce the risk of rock burst.

[0106] Specifically, the deep-buried hard rock tunnel TBM rapid excavation device provided in the embodiment of this scheme is used to execute the deep-buried hard rock tunnel TBM rapid excavation method mentioned above in this scheme. Its implementation method is consistent with the implementation method of the deep-buried hard rock tunnel TBM rapid excavation method provided in this scheme, and can achieve the same beneficial effects, which will not be repeated here.

[0107] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the technical solutions of the present invention, and are not intended to limit the specific implementation methods of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the claims of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for rapid excavation of a deep hard rock tunnel using a TBM, characterized in that: The method comprises the following steps: S1: Use a horizontal directional drill to conduct advance drilling and monitor the torque of the drill bit in real time; S2: Uses a machine learning and deep learning algorithm combining convolutional neural networks and support vector machines to perform stratigraphic image recognition on core samples transported from advance drilling. S3: Implement precise explosive placement and blasting operations based on drill bit torque monitoring and formation image recognition results; S4: Perform TBM excavation operations.

2. A method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 1, characterized in that: The step S2 comprises the following steps: S21: Capture images of core samples transported from advance drilling and perform data preprocessing on the images; S22: Using convolutional neural network to extract features from preprocessed image data; S23: Input the extracted features into the support vector machine for classification; S24: Compare the classification results with the stratigraphic database to identify different stratigraphic types and generate a geological report.

3. A method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 2, characterized in that: The step S21 includes: performing image enhancement using data enhancement technology and performing image normalization processing.

4. A method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 3, characterized in that: The normalization formula is: Among them, l rdw is the original image pixel value, min(I) and max(I) are the minimum and maximum values ​​of the image pixels respectively, I norm is the normalized image pixel value.

5. The method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 2, characterized in that: In step S22, local features of the image are extracted by a convolution operation, and the formula of the convolution operation is: Where f(x,y) represents the convolution result, I(x,y) is the original pixel value of the image, K(i,j) is the convolution kernel, k is the size of the convolution kernel, and x and y are the image coordinates.

6. A method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 2, characterized in that: In step S23, a classification rule is formulated through a classification decision function, and then a suitable kernel function is selected to distinguish different stratum types; wherein the classification decision function formula is: Among them, x is the sample to be classified, x i is the support vector, α i is the Lagrange multiplier, y i is the sample label, K(x i ,x) is the kernel function, b is the bias term, and sign is the sign function.

7. A method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 6, characterized in that: The kernel function is a radial basis function, which is used to capture complex features in the image and distinguish different stratum types. The formula of the radial basis function is: Among them, σ represents a hyperparameter; |x i -x j | represents sample x i and x j The Euclidean distance between .

8. The method for rapid excavation of a deep hard rock tunnel using a TBM according to claim 2, characterized in that: In the step S24, it includes: S241: Designing a stratum database according to information of different stratum types stored in the database; S242: Initiating a query to the stratum database based on the classification result label and the feature extraction result, and retrieving the stratum database to obtain stratum information related to the classification result; S243: The retrieved stratigraphic information is compared with the stratigraphic image recognition result. If they are consistent, the classification result of the stratigraphic image is correct. If they are inconsistent, the feedback mechanism is triggered to perform warning and feedback operations, trigger manual inspection or further analysis, and finally generate a final geological report.

9. A method for rapid excavation of a deep hard rock tunnel using a TBM according to any one of claims 1 to 8, characterized in that: In step S4, the operating status of the TBM is also monitored in real time, the working data of the TBM is collected in real time through sensors, and a visual monitoring platform is established to display all data in real time. At the same time, the excavation parameters can be adjusted according to the monitored excavation operation conditions.

10. A deep hard rock tunnel TBM rapid excavation device based on the deep hard rock tunnel TBM rapid excavation method according to any one of claims 1 to 9, characterized in that: The device comprises: Drilling module: used for advanced drilling using a horizontal directional drill and real-time monitoring of the drill bit torque; Image recognition module: This module uses a machine learning and deep learning algorithm that combines a convolutional neural network with a support vector machine to perform formation image recognition on core samples transported from advance drilling. Blasting module: used to implement precise explosive placement and blasting operations based on drill bit torque monitoring and formation image recognition results; Tunneling operation module: used for TBM tunneling operations.

Images